Ticket #4992: llvm-primcall-register-trashing.patch

File llvm-primcall-register-trashing.patch, 93.5 KB (added by scpmw, 9 years ago)
Line 
11 patch for repository http://darcs.haskell.org/ghc:
2
3Tue Mar  1 18:55:59 GMT 2011  scpmw@leeds.ac.uk
4  * LLVM code gen: Do not trash registers for primitive calls
5 
6  Primitive calls won't do anything with global registers, trashing them just confuses LLVM and produces crashing programs.
7
8New patches:
9
10[LLVM code gen: Do not trash registers for primitive calls
11scpmw@leeds.ac.uk**20110301185559
12 Ignore-this: 10e519ef64025e63bef54b14572210fd
13 
14 Primitive calls won't do anything with global registers, trashing them just confuses LLVM and produces crashing programs.
15] hunk ./compiler/llvmGen/LlvmCodeGen/CodeGen.hs 304
16        need are restored from the Cmm local var and the ones we don't need
17        are fine to be trashed.
18     -}
19-    let trashStmts = concatOL $ map trashReg activeStgRegs
20+    let trashStmts | cconv == PrimCallConv = nilOL
21+                   | True                  = concatOL $ map trashReg activeStgRegs
22             where trashReg r =
23                     let reg   = lmGlobalRegVar r
24                         ty    = (pLower . getVarType) reg
25
26Context:
27
28[Restore SPECIALISE INSTANCE pragmas
29simonpj@microsoft.com**20110222173112
30 Ignore-this: 4c0a8fab4f9e619d988beb8b5269474f
31 
32 For some reason I'd commented out this code, which
33 meant that we weren't getting the the goodness of
34 the SPECIALISE INSTANCE pragamas in GHC.Real. And
35 that in turn killed performance in nofib 'power'
36]
37[GHC.Prim.threadStatus# now returns the cap number, and the value of TSO_LOCKED
38Simon Marlow <marlowsd@gmail.com>**20110301101025
39 Ignore-this: 49548cffd04bf6fdd3071158d9865034
40]
41[Improve GHCi line numbers in errors
42Ian Lynagh <igloo@earth.li>**20110227172012
43 Ignore-this: 52c2c2f7c83207e8c7b36fab6ebe52d8
44 When running commands from the user (as opposed to from a file), reset
45 the line number to 1 at the start of each command.
46]
47[:script file scripts in GHCi #1363
48Vivian McPhail <haskell.vivian.mcphail@gmail.com>**20110226073133
49 This patch adds the script command in GHCi
50 
51 A file is read and executed as a series of GHCi commands.
52 
53 Execution terminates on the first error.  The filename and
54 line number are included in the error.
55]
56[Turn off split objects on Darwin if XCode < 3.2 (#4013)
57Ian Lynagh <igloo@earth.li>**20110225184358
58 Ignore-this: 3e894673649fee957df2413ca82b1c99
59]
60[libffi: backport incorrect detection of selinux
61Sergei Trofimovich <slyfox@community.haskell.org>**20110208212140
62 Ignore-this: 31bdddb5683acae2bffa3e97effaefea
63 
64 This patch unbreaks ghci on GRSEC kernels hardened with
65 TPE (Trusted Path Execution) protection.
66 
67 TPE forbids mmap('rwx') files opened for writes:
68     fd = open (a_file_in_tmp, O_RDWR);
69     mmap (...,               PROT_READ | PROT_WRITE | PROT_EXEC, fd);
70 
71 while allows anonymous RWX mappings:
72     mmap (...MAP_ANONYMOUS , PROT_READ | PROT_WRITE | PROT_EXEC, -1);
73 
74 Thanks to klondike for finding it out.
75 
76 The result of a horrible typo.
77 
78 (unreleased yet) upstream also has the patch:
79 
80 http://github.com/atgreen/libffi/commit/eaf444eabc4c78703c0f98ac0197b1619c1b1bef
81]
82[Generate the OS X installer from a bindist, rather than from a source tree
83Ian Lynagh <igloo@earth.li>**20110223170016]
84[Use -h rather than -soname; fixes dynlibs on Solaris 10; trac #4973
85Ian Lynagh <igloo@earth.li>**20110222152656]
86[Fix another fundep error (fixes Trac #4969)
87simonpj@microsoft.com**20110221153239
88 Ignore-this: 9308a7c55346c9eaeaa0818903c8d67b
89 
90 If I had a pound for every hour Dimitrios and I have spent
91 making functional dependencies work right, we'd be rich!
92 
93 We had stupidly caused a 'wanted' to be rewritten by a 'derived', with
94 resulting abject failure.  As well as fixing the bug, this patch
95 refactors some more, adds useful assert and comments.
96]
97[New codegen: GC calling convention must use registers.
98Edward Z. Yang <ezyang@mit.edu>**20110218011745
99 Ignore-this: 13cdd38bb5cea0d244861f01f0004dee
100 
101 Previously, on register-deficient architectures like x86-32,
102 the new code generator would emit code for calls to stg_gc_l1,
103 stg_gc_d1 and stg_gc_f1 that pushed their single argument on
104 to the stack, while the functions themselves expected the
105 argument to live in L1, D1 and F1 (respectively).  This was
106 because cmmCall with the GC calling convention allocated real
107 registers, not virtual registers.
108 
109 This patch modifies the code for assigning registers/stack slots
110 to use the right calling convention for GC and adds an assertion
111 to ensure it did it properly.
112]
113[Added a VECTORISE pragma
114Manuel M T Chakravarty <chak@cse.unsw.edu.au>**20110220105032
115 Ignore-this: e3ed0e7caa45418452858a5a5937264
116 - Added a pragma {-# VECTORISE var = exp #-} that prevents
117   the vectoriser from vectorising the definition of 'var'.
118   Instead it uses the binding '$v_var = exp' to vectorise
119   'var'.  The vectoriser checks that the Core type of 'exp'
120   matches the vectorised Core type of 'var'.  (It would be
121   quite complicated to perform that check in the type checker
122   as the vectorisation of a type needs the state of the VM
123   monad.)
124 - Added parts of a related VECTORISE SCALAR pragma
125 - Documented -ddump-vect
126 - Added -ddump-vt-trace
127 - Some clean up
128]
129[Reenable object splitting on Darwin, now #4013 appears to be fixed
130Ian Lynagh <igloo@earth.li>**20110219191409
131 Ignore-this: e39e46beab8c8f2dc3369d5ec98610f2
132]
133[Clarify the "object splitting" variable names
134Ian Lynagh <igloo@earth.li>**20110219161508]
135[Fix #4867 (updated; corrects address calculation)
136gwright@antiope.com**20110214150924
137 Ignore-this: b68bcf1262f3507e1b9a7459139054bc
138 
139 This is a corrected fix for ticket #4867, "ghci displays negative floats
140 incorrectly".  The previous patch sometimes gave incorrect offset to values
141 in the __const section of the __TEXT segment.  The new patch arranges a zero
142 fixup for non-external, not-global offset table signed relocations.  This
143 is apparently what is required, though documentation on this point is scarce.
144 
145 With this change Doubles are negated properly, because the sign bit mask
146 is loaded from the correct offset.  This was tested both on HEAD and the 7.0
147 branch.
148 
149]
150[Scale columns in cost-centre-stack report to their contents
151Max Bolingbroke <batterseapower@hotmail.com>**20110216161340
152 Ignore-this: b7138527bb8c139ed554b9e6c9a9f37
153]
154[Change typechecker-trace output slightly
155simonpj@microsoft.com**20110217175904
156 Ignore-this: 4bf1c9dfa41c598765cd082a35bf193b
157]
158[Comments only
159simonpj@microsoft.com**20110217175840
160 Ignore-this: b9ea9687b41bfc062137a3ad6fafa3a6
161]
162[Fix Trac #4966
163simonpj@microsoft.com**20110217141000
164 Ignore-this: d4caaf910dfb86b1e469c5da3fe9cde2
165 
166 This is just a program that exploits overlapping
167 instances in a delicate way. The fix makes GHC
168 a bit more friendly towards such programs.
169 
170 See Note [Overlap and deriving] in TcSimplify
171]
172[Use "on the spot" solving for fundeps
173simonpj@microsoft.com**20110217140921
174 Ignore-this: 99290ef5538819053af708f5f9464488
175 
176 When we spot an equality arising from a functional dependency,
177 we now use that equality (a "wanted") to rewrite the work-item
178 constraint right away.  This avoids two dangers
179 
180  Danger 1: If we send the original constraint on down the pipeline
181            it may react with an instance declaration, and in delicate
182           situations (when a Given overlaps with an instance) that
183           may produce new insoluble goals: see Trac #4952
184 
185  Danger 2: If we don't rewrite the constraint, it may re-react
186            with the same thing later, and produce the same equality
187            again --> termination worries.
188 
189 To achieve this required some refactoring of FunDeps.lhs (nicer
190 now!). 
191 
192 This patch also contains a couple of unrelated improvements
193 
194 * A bad bug in TcSMonad.nestImplicTcS whereby the Tcs tyvars
195   of an outer implication were not untouchable inside
196 
197 * Improved logging machinery for the type constraint solver;
198   use -ddump-cs-trace (probably with a wider default line width
199   -dppr-cols=200 or something)
200]
201[Increase exprIsDupable threshold a bit
202simonpj@microsoft.com**20110215143921
203 Ignore-this: cf5a73026aabaa9f2ca93d0d338a6f56
204 
205 Now that exprIsDupable is less aggressive, test MethSharing wasn't
206 doing enough inlining.  Increasing the threshold fixes the problem
207 but the real fix is in Trac #4960.
208]
209[Ensure exprIsCheap/exprIsExpandable deal with Cast properly
210simonpj@microsoft.com**20110215143655
211 Ignore-this: 13bd5540b048de02737070bfa66f1a71
212 
213 This bug was causing a Lint error on the stable branch. For some
214 reason it doesn't show up in HEAD, but it's still worth fixing.
215 
216 The point is that ((f `cast` co) a) is cheap if f has arity>1.
217 This was being gratuitously missed before.
218]
219[Do not treat absentError specially
220simonpj@microsoft.com**20110215143504
221 Ignore-this: 3a8f6d6eae17b7360f58de86ec934cde
222 
223 (This was part of an experiment I abandoned.)
224]
225[Fix exprIsDupable
226simonpj@microsoft.com**20110214111512
227 Ignore-this: 2b800410f8103590da167ae24e7b242d
228 
229 It turns out that exprIsDupable would return True for an expression of
230 *arbitrary* size, provide it was a nested bunch of applications in
231 which no function had more than three arguments.  That was never the
232 intention, and could give rise to massive code duplication. 
233 
234 This patch makes it much less aggressive.
235]
236[Better case-of-case transformation
237simonpj@microsoft.com**20110214111151
238 Ignore-this: 32360d2c7703d772c7a5ad84b019bf87
239 
240 The ThunkSplitting idea in WorkWrap wasn't working at all,
241 leading to Trac #4957.  The culprit is really the simplifier
242 which was combining the wrong case continuations. See
243 Note [Fusing case continuations] in Simplify.
244]
245[Cleaned up Expr and Vectorise
246keller@.cse.unsw.edu.au**20110215014434
247 Ignore-this: 11b78040a7d16dc6e5ce7950c73bf8a9
248]
249[Fixed two syntax errors
250keller@.cse.unsw.edu.au**20110214020531
251 Ignore-this: 1abeebe0171a4262ceacee89adc9246e
252]
253[Handling of recursive scalar functions in isScalarLam
254keller@cse.unsw.edu.au**20110214002945
255 Ignore-this: db522152971fdcc9c74124d06ef785bd
256]
257[Added handling of non-recursive module global functions to isScalar check
258keller@cse.unsw.edu.au**20110209042855
259 Ignore-this: feae370a9bcc30371ab62f3700471739
260]
261[
262keller@cse.unsw.edu.au**20110202051408
263 Ignore-this: 7e0e1fcd40315da9abe80e2c2eda877f
264]
265[Removed minor bug
266keller@cse.unsw.edu.au**20110202040826
267 Ignore-this: 1abdae6a0f39b2ebd002f2587b5d1f53
268]
269[added handling of data constructors to vectLam
270keller@cse.unsw.edu.au**20110201042807
271 Ignore-this: 511fba94f926bb3a1088b21e36cc1f8c
272]
273[pruneSparkQueue: check for tagged pointers
274Simon Marlow <marlowsd@gmail.com>**20110214123858
275 Ignore-this: 719f77302b0dff3ba97b66a0d75b2482
276 This was a bug in 6.12.3.  I think the problem no longer occurs due to
277 the way sparks are treated as weak pointers, but it doesn't hurt to
278 test for tagged pointers anyway: better to do the test than have a
279 subtle invariant.
280]
281[Fix Trac #4953: local let binders can have IdInfo with free names
282simonpj@microsoft.com**20110214140334
283 Ignore-this: c8e5dbc6f0270e45d9cf2edee34d0354
284 
285 Local let binders in IfaceExpr never used to have unfoldings,
286 but lately they can (becuase they can have an INLINE pragma).
287 We must take account of the variables mentioned in the pragma
288 when computing the fingerprint.
289]
290[Comments only
291simonpj@microsoft.com**20110214090703
292 Ignore-this: bf950e91e0c984186fea1100aa15d00c
293]
294[LLVM: Huge improvement to mangler speed.
295David Terei <davidterei@gmail.com>**20110213014406
296 Ignore-this: 4eeaa572dfe956c990895154bd942bb2
297 
298 The old llvm mangler was horrible! Very slow
299 due to bad design and code. New version is
300 linear complexity as it should be and far
301 lower coefficients. This fixes trac 4838.
302]
303[Fix platform detection in bindists
304Ian Lynagh <igloo@earth.li>**20110211184244
305 In a bindist, we generate files like the hsc2hs wrapper.
306 This means we need to have the right values for the variables like
307 CONF_GCC_LINKER_OPTS_STAGE1 which in turn means we need to know what
308 platform we're on.
309]
310[New plan: push unsolved wanteds inwards
311simonpj@microsoft.com**20110211174058
312 Ignore-this: ed40762e260dab75b5e51c696f9965fa
313 
314 This fixes Trac #4935.  See Note [Preparing inert set for implications].
315 Lots of comments, but not a lot of code is changed!
316]
317[Remove unnecessary import, plus white space
318simonpj@microsoft.com**20110211173925
319 Ignore-this: 963f455c3c8aee49009b5d5a02f835ac
320]
321[Fix small but egregious error: using un-zonked constraints in simplifyRule
322simonpj@microsoft.com**20110211173835
323 Ignore-this: 238586e420dbbb1be7f6368117cf6280
324 
325 This resulted in double unifications.  Fix is trivial.
326]
327[makeSolvedByInst is only called on wanteds
328simonpj@microsoft.com**20110211173415
329 Ignore-this: 36e6201ab59a082e6dc38e56bea99e29
330 
331 This patch just adds an assert error.
332]
333[Enable DTrace on Solaris; based on a patch from Karel Gardas
334Ian Lynagh <igloo@earth.li>**20110210155217
335 Ignore-this: 93eaf0e06c721c80c175aaee9a113e6
336]
337[Use DTrace whenever it's available
338Ian Lynagh <igloo@earth.li>**20110210153300
339 Ignore-this: 111c72bf20d6eaafd3e488196a89b2c
340 Now that we've stopped trying to support 64bit OS X 10.5, the DTrace
341 problems there don't matter.
342]
343[replace C++ comments with C comments (Solaris' DTrace fails on C++ comments)
344Karel Gardas <karel.gardas@centrum.cz>**20110112051829
345 Ignore-this: c229292227c7e2b512daf9129cb66aeb
346]
347[Fix #4867, ghci displays negative floats incorrectly
348gwright@antiope.com**20110209222423
349 Ignore-this: 1b3279fa5f6c4849ed6311275b1a466a
350 
351 This patch fixes the erroneous relocations that caused
352 the bug in ticket #4867.  External addresses and global
353 offset table entries were relocated correctly, but all other
354 relocations were incorrectly calculated.  This caused, for
355 example, bad references to constants stored in the __const
356 section of the __TEXT segment.
357 
358 This bug only affected OS X on 64-bit platforms.
359 
360]
361[Fix Array sizeof primops to use the correct offset (which happens to be 0, so it worked before anyway). Makes us more future-proof, at least
362Daniel Peebles <pumpkingod@gmail.com>**20110201063017
363 Ignore-this: 8e79c3f6f80c81b4160a31e80e4ed29d
364]
365[Add sizeof(Mutable)Array# primitives
366Daniel Peebles <pumpkingod@gmail.com>**20110126051554
367 Ignore-this: ae17d94dbb86d6e1ffa0a489da842f78
368]
369[fix TRY_ACQUIRE_LOCK on Windows.
370Simon Marlow <marlowsd@gmail.com>**20110210150035
371 Ignore-this: b48713585b6d65020bb42f952c2c54ac
372]
373[constant fold  (a + N) - M  and  (a - N) + M
374Simon Marlow <marlowsd@gmail.com>**20110210115608
375 Ignore-this: 312c33d5ff7d4288be8ab0a5c6939c0c
376]
377[Recursively call cmmMachOpFold on divides that we turned into shifts
378Simon Marlow <marlowsd@gmail.com>**20110208104345
379 Ignore-this: bca0db454127c5f2b64b81d6200fc000
380 There might be more simplification to do.
381]
382[Add unboxed tuple support to Template Haskell
383Ian Lynagh <igloo@earth.li>**20110210134528
384 Ignore-this: cf946570a9d16016debf8bccd21b2c79
385]
386[Allow TH brackets to contain things of any kind
387Ian Lynagh <igloo@earth.li>**20110209184459
388 Ignore-this: b23f53d201abc874cf1f15a4eddb3abb
389 You can now quasi-quote things with unboxed types, and unboxed tuples.
390]
391[Simpify constraints from a TH bracket eagerly
392simonpj@microsoft.com**20110209175003
393 Ignore-this: b341ea3d235af1b2e617107f238ae1d6
394 
395 See Trac #4949, where having a TH bracket implication
396 was messing things up.  Better to get rid of it right away.
397]
398[Typo in comment
399simonpj@microsoft.com**20110209174914
400 Ignore-this: 3590bf097a566c6f590a37cda2be9b4e
401]
402[Call the final build system phase "final" rather than ""
403Ian Lynagh <igloo@earth.li>**20110207142046
404 Ignore-this: cc87cf202cff1f1d8b105268dacdd63f
405]
406[Fix bug introduced in "Implement fuzzy matching for the Finder"
407Simon Marlow <marlowsd@gmail.com>**20110208090121
408 Ignore-this: 53e61080e7d7fb28f9187629fa20746a
409 The finder was reporting a hidden package when it meant a hidden
410 module, and vice versa (looks like a typo).
411]
412[Fix Trac #4945: another SpecConstr infelicity
413simonpj@microsoft.com**20110207102537
414 Ignore-this: c3ffbb640cdbbab32758e6130ae803bc
415 
416 Well, more a plain bug really, which led to SpecConstr
417 missing some obvious opportunities for specialisation.
418 
419 Thanks to Max Bolingbroke for spotting this.
420]
421[add missing initialisation of ws->todo_large_objects
422Simon Marlow <marlowsd@gmail.com>**20110204093148
423 Ignore-this: dc9a28f85aff97e0896d212d7b21ae30
424 Found-by: Valgrind.  Thanks Julian!
425]
426[Add -XNondecreasingIndentation to -XHaskell98 for backwards compatibility.
427Simon Marlow <marlowsd@gmail.com>**20110204084226
428 Ignore-this: 42f7ef8bfbfb8d58f61afa217af3ffea
429 The final straw was when I learned today that Happy broke.
430]
431[only the GHC repo is in git for now; add hoopl
432Simon Marlow <marlowsd@gmail.com>**20110203205705
433 Ignore-this: 1cf90e7e5b83aabef74ec5e24496464c
434]
435[Fix typo in SpecConstr that made it not work at all
436simonpj@microsoft.com**20110203172756
437 Ignore-this: b550d5c5b73ed13709ee2938c80a750f
438 
439 There was a terrible typo in this patch; I wrote "env"
440 instead of "env1".
441 
442    Mon Jan 31 11:35:29 GMT 2011  simonpj@microsoft.com
443      * Improve Simplifier and SpecConstr behaviour
444 
445 Anyway, this fix is essential to make it work properly.
446 Thanks to Max for spotting the problem (again).
447]
448[fix compacting GC
449Simon Marlow <marlowsd@gmail.com>**20110202170036
450 Ignore-this: e78c99b318586a7fccc2a8e36d9fbf88
451]
452[fix warning
453Simon Marlow <marlowsd@gmail.com>**20110202160415
454 Ignore-this: 99f65d20b38cce971b0eda6c53eab8d3
455]
456[GC refactoring and cleanup
457Simon Marlow <marlowsd@gmail.com>**20110202154955
458 Ignore-this: 96b5b5ec97d49e69617d0007ee7fe804
459 Now we keep any partially-full blocks in the gc_thread[] structs after
460 each GC, rather than moving them to the generation.  This should give
461 us slightly better locality (though I wasn't able to measure any
462 difference).
463 
464 Also in this patch: better sanity checking with THREADED.
465]
466[avoid adding HPC ticks to arrow constructs (fixes #1333)
467Ross Paterson <ross@soi.city.ac.uk>**20110202211425
468 Ignore-this: 2938850ebbb53d1bc6bf0399f68dd8e5
469]
470[Fix the profiling build
471Simon Marlow <marlowsd@gmail.com>**20110202132257
472 Ignore-this: cdf6de609c7eee47bd6c9e957276f12b
473]
474[A small GC optimisation
475Simon Marlow <marlowsd@gmail.com>**20110202123049
476 Ignore-this: 4119e33b0f40787fd9339ad1104b3b9e
477 Store the *number* of the destination generation in the Bdescr struct,
478 so that in evacuate() we don't have to deref gen to get it.
479 This is another improvement ported over from my GC branch.
480]
481[scheduleProcessInbox: use non-blocking acquire, and take the whole queue
482Simon Marlow <marlowsd@gmail.com>**20110202114907
483 Ignore-this: 12020d2751d355d1c006697351223d99
484 This is an improvement from my GC branch, that helps performance for
485 intensive message-passing communication between Capabilities.
486]
487[do a bit of by-hand CSE
488Simon Marlow <marlowsd@gmail.com>**20110202114417
489 Ignore-this: c0c90cd767c74f3eee8b7f8cbc08dfa0
490]
491[add a const
492Simon Marlow <marlowsd@gmail.com>**20110202114345
493 Ignore-this: d12300d69c91d7187aa1dd83a4a13ff9
494]
495[add TRY_ACQUIRE_LOCK()
496Simon Marlow <marlowsd@gmail.com>**20110202113242
497 Ignore-this: face9da80ce407b9013d5f73bb65c34d
498]
499[Remove the per-generation mutable lists
500Simon Marlow <marlowsd@gmail.com>**20110202112646
501 Ignore-this: 3e0cacbc8c8b6ddf7005d25b593d3357
502 Now that we use the per-capability mutable lists exclusively.
503]
504[+RTS -qw hasn't done anything since 7.0.1; remove the implementation & docs
505Simon Marlow <marlowsd@gmail.com>**20110201163727
506 Ignore-this: e6c6ba6b8a119d87efcd79310b4fb5d2
507 It is still (silently) accepted for backwards compatibility.
508]
509[comments
510Simon Marlow <marlowsd@gmail.com>**20110201085830
511 Ignore-this: 5a70e58a48aa60e5ed7afc6e908d150a
512]
513[Annotate thread stop events with the owner of the black hole
514Simon Marlow <marlowsd@gmail.com>**20110127164226
515 Ignore-this: a60cbe5cc50da911d58020775c513ed0
516 
517 So we can now get these in ThreadScope:
518 
519   19487000: cap 1: stopping thread 6 (blocked on black hole owned by thread 4)
520 
521 Note: needs an update to ghc-events.  Older ThreadScopes will just
522 ignore the new information.
523]
524[update debugging code for fragmentation
525Simon Marlow <marlowsd@gmail.com>**20110125111011
526 Ignore-this: a98cd31d2e48ae9bdc52f9d96424ce39
527]
528[Fix type checker error message
529simonpj@microsoft.com**20110201122920
530 Ignore-this: 7369cc5f8dae3d81621f580a8ddaf41e
531 
532 See Trac #4940. We had a message
533      The lambda expression `\ x -> x' has one argument one argument,
534 repeating the "one argument" part.  Easy fix.
535]
536[Some refactoring of SpecConstr
537simonpj@microsoft.com**20110201122841
538 Ignore-this: c2966091564a9ca4ceb27a9596d36b7d
539 
540 This was originally to improve the case when SpecConstr generated a
541 function with an unused argument (see Trac #4941), but I ended up
542 giving up on that.  But the refactoring is still an improvement.
543 
544 In particular I got rid of BothOcc, which was unused.
545]
546[Don't make join points when the case has only one non-bottom alternative
547simonpj@microsoft.com**20110201122637
548 Ignore-this: 333b9b62debbfc1338530ef4e710ccbb
549 
550 This fixes Trac #4930.  See Note [Bottom alternatives] in Simplify.lhs
551]
552[Improve Simplifier and SpecConstr behaviour
553simonpj@microsoft.com**20110131113529
554 Ignore-this: e5b96c97cee0950e558ddf15178bb6c9
555 
556 Trac #4908 identified a case where SpecConstr wasn't "seeing" a
557 specialisation it should easily get.  The solution was simple: see
558 Note [Add scrutinee to ValueEnv too] in SpecConstr.
559 
560 Then it turned out that there was an exactly analogous infelicity in
561 the mighty Simplifer too; see Note [Add unfolding for scrutinee] in
562 Simplify. This fix is good for Simplify even in the absence of the
563 SpecConstr change.  (It arose when I moved the binder- swap stuff to
564 OccAnall, not realising that it *remains* valuable to record info
565 about the scrutinee of a case expression.  The Note says why.
566 
567 Together these two changes are unconditionally good.  Better
568 simplification, better specialisation. Thank you Max.
569]
570[fix warning
571Simon Marlow <marlowsd@gmail.com>**20110131135951
572 Ignore-this: e893d9bfbabf1601133a1e09c50b908
573]
574[32-bit fix
575Simon Marlow <marlowsd@gmail.com>**20101013154200
576 Ignore-this: 7508977c263ed2cec321b40a8b5772a
577]
578[update to mingw gcc 4.5.2
579Simon Marlow <marlowsd@gmail.com>**20110119135053
580 Ignore-this: b9d5cb736a48a0adc5e35eb8a0c191cd
581]
582[count fizzled and GC'd sparks separately
583Simon Marlow <marlowsd@gmail.com>**20101111132727
584 Ignore-this: 4cb4b759aed06659b46cdf76e791e5c9
585]
586[count "dud" sparks (expressions that were already evaluated when sparked)
587Simon Marlow <marlowsd@gmail.com>**20101101124143
588 Ignore-this: ca94824c0e75da0b3688300e7285c7e6
589]
590[fix some shutdown memory leaks
591Simon Marlow <marlowsd@gmail.com>**20100820093133
592 Ignore-this: 3e7b80b5f4846d6c56319c150895953d
593]
594[fix DEBUG build
595Simon Marlow <marlowsd@gmail.com>**20110131123433
596 Ignore-this: f2e009eaa66a14a7c8ec6acc7a4bbdb1
597]
598[Fix formatting glitch in documentation
599simonpj@microsoft.com**20110128115400
600 Ignore-this: 6c410ed19956feac7e0cf68bb40b40b1
601]
602[Fix warnings
603Simon Marlow <marlowsd@gmail.com>**20110128103639
604 Ignore-this: aa7f2c9f9b91f9dabc7b5d5ea26121fd
605]
606[Merge in new code generator branch.
607Simon Marlow <marlowsd@gmail.com>**20110124121650
608 Ignore-this: 7762f21082cb84ec94daaeefd70f5ef2
609 This changes the new code generator to make use of the Hoopl package
610 for dataflow analysis.  Hoopl is a new boot package, and is maintained
611 in a separate upstream git repository (as usual, GHC has its own
612 lagging darcs mirror in http://darcs.haskell.org/packages/hoopl).
613 
614 During this merge I squashed recent history into one patch.  I tried
615 to rebase, but the history had some internal conflicts of its own
616 which made rebase extremely confusing, so I gave up. The history I
617 squashed was:
618 
619   - Update new codegen to work with latest Hoopl
620   - Add some notes on new code gen to cmm-notes
621   - Enable Hoopl lag package.
622   - Add SPJ note to cmm-notes
623   - Improve GC calls on new code generator.
624 
625 Work in this branch was done by:
626    - Milan Straka <fox@ucw.cz>
627    - John Dias <dias@cs.tufts.edu>
628    - David Terei <davidterei@gmail.com>
629 
630 Edward Z. Yang <ezyang@mit.edu> merged in further changes from GHC HEAD
631 and fixed a few bugs.
632]
633[Fix an egregious strictness analyser bug (Trac #4924)
634simonpj@microsoft.com**20110128080748
635 Ignore-this: 3bf533c3d30b45a8e78b1fec3d9634f
636 
637 The "virgin" flag was being threaded rather than treated
638 like an environment.  As a result, the second and subsequent
639 recursive definitions in a module were not getting a
640 correctly-initialised fixpoint loop, causing much worse
641 strictness analysis results.  Indeed the symptoms in
642 Trac #4924 were quite bizarre.
643 
644 Anyway, it's easily fixed.  Merge to stable branch.
645]
646[Refine incomplete-pattern checks (Trac #4905)
647simonpj@microsoft.com**20110127131304
648 Ignore-this: cf2e0852f20d1cadc6a2cba4272838f6
649 
650 The changes are:
651 
652 * New flag -fwarn-incomplete-uni-patterns, which checks for
653   incomplete patterns in (a) lambdas, (b) pattern bindings
654 
655 * New flag is not implied by -W or -Wall (too noisy; and many
656   libraries use incomplete pattern bindings)
657 
658 * Actually do the incomplete-pattern check for pattern bindings
659   (previously simply omitted)
660 
661 * Documentation for new flag
662]
663[Fix "make 1" etc following the build system changes
664Ian Lynagh <igloo@earth.li>**20110127001739
665 Ignore-this: 7ae0a41f2753d7740569f362a97ea5fb
666 The logic is now in mk/compiler-ghc.mk rather than being duplicated in
667 ghc/Makefile and compiler/Makefile.
668]
669[Fix vectorisation of recursive types
670Roman Leshchinskiy <rl@cse.unsw.edu.au>**20110126231843
671 Ignore-this: 983fc42a659be2e085da9b16f994aa2e
672]
673[Fix dependencies among specialisations for imported Ids
674simonpj@microsoft.com**20110126172112
675 Ignore-this: 364e09c11affe7bfe8f1b934ea28bbb6
676 
677 This was a subtle one (Trac #4903).  See
678   Note [Glom the bindings if imported functions are specialised]
679 in Speclialise.
680 
681 Fundamentally, a specialised binding for an imported Id was being
682 declared non-recursive, whereas in fact it can become recursive
683 via a RULE.  Once it's specified non-recurive the OccAnal pass
684 treats that as gospel -- and that in turn led to infinite inlining.
685 
686 Easily fixed by glomming all the specialised bindings in a Rec;
687 now the OccAnal will sort them out correctly.
688]
689[Fix bug in roughTopNames
690simonpj@microsoft.com**20110126171803
691 Ignore-this: eca8b144162f1bd94e2ccb433bca1e02
692 
693 roughTopNames was returning a name that in fact might be
694 "looked though" by the rule matcher. Result: a rule
695 that should match was being pre-emptively discarded.
696 
697 See Note [Care with roughTopName].
698 
699 Fixes a bug noticed by Pedro (Trac #4918).
700]
701[Comments only, plus a tiny bit of debug printing
702simonpj@microsoft.com**20110126171255
703 Ignore-this: f84364b2b90fc860e9289dd40d0395ac
704]
705[Comments only
706simonpj@microsoft.com**20110126171235
707 Ignore-this: 79059977f82aaac7f9714ad09e820ea9
708]
709[Look through type synonyms when computing orphans
710simonpj@microsoft.com**20110126171229
711 Ignore-this: 6dfc45dae3a94cdb0022b2d21d6e09f6
712 
713 I renamed functions tyClsNamesOfTypes to oprhNamesOfType,
714 because it's only used in that capacity, and we therefore
715 want to look through type synonyms.  Similarly exprOrphNames.
716 
717 This fixes Trac #4912.
718]
719[Bleat a bit more informatively in unionLists
720simonpj@microsoft.com**20110126171030
721 Ignore-this: 80b276aa3d7971c6d7802b5f6b522d2e
722]
723[Keep separate linker flags, for when we want to link with gcc or ld
724Ian Lynagh <igloo@earth.li>**20110124233121]
725[Fix validate on OS X 64
726Ian Lynagh <igloo@earth.li>**20110124183618]
727[Split main/GHC into GHC and GhcMake
728simonpj@microsoft.com**20110125161632
729 Ignore-this: 502ea034de77ecd81173161836d78287
730 
731 There are two things going on in main/GHC.hs.
732   * It's the root module of the GHC package
733   * It contains lots of stuff for --make
734 It is also gigantic (2.7k lines)
735 
736 This patch splits it into two
737   * GHC.hs is the root module for the GHC package
738       (1.3k lines)
739   * GhcMake.hs contains the stuff for --make
740       (1.4k lines)
741 
742 Happily the functional split divided it almost
743 exactly in half.
744 
745 This is a pure refactoring.  There should be no
746 behavioural change.
747]
748[Comments only
749simonpj@microsoft.com**20110125131115
750 Ignore-this: 7ec4e97a481d06894de940aba59c575d
751]
752[Fix Trac #3717 by making exprOkForSpeculation a bit cleverer
753simonpj@microsoft.com**20110125110525
754 Ignore-this: 13b606b05da69c29bf53aaf408fd602
755 
756 The main change here is to do with dropping redundant seqs.
757 See Note [exprOkForSpeculation: case expressions] in CoreUtils.
758]
759[Improve dataToTag# magic
760simonpj@microsoft.com**20110125110418
761 Ignore-this: 11fdb265e030dec4d5b13ed6b16c9761
762 
763 dataToTag# is a bit unsatisfactory because it requires
764 its argument to be evaluated, and we don't have a good
765 way to enforce that. This patch adds some comments, and
766 makes exprOkForSpeculation a bit less picky in the case
767 of dataToTag# (since the argument may, in fact, not be
768 eval'd).
769]
770[Fix Trac #4917: try a bit harder to unify on-the-fly
771simonpj@microsoft.com**20110125110112
772 Ignore-this: e96e0a19ab8517d4ba648efe91f6b379
773 
774 This is generally a modest improvement but, more important,
775 it fixes a "unify-under-forall" problem.  See Note [Avoid deferring].
776 
777 There's still a lurking unsatisfactory-ness in that we can't
778 defer arbitrary constraints that are trapped under a forall.
779]
780[DPH options updated
781Manuel M T Chakravarty <chak@cse.unsw.edu.au>**20110124043617
782 Ignore-this: 6b7d2949b75f9c923f279c1178d2d042
783 - '-Odph' is now '-O2 -fsimplifier-phases=3 -fsimplifier-iterations=20'
784 - The new option '-fdph-none' is the default; it indicates that no DPH
785   backend is selected and is the only valid option if the DPH libraries
786   are not installed.  If vectorisation is attempted with -fdph-none a
787   suitable error message is generated.
788 - Hence, '-fdph-par' (or '-fdph-seq') needs to be explicitly selected
789   when using vectorisation and when linking vectorised code.  (There
790   seems to be no elegant way to avoid that.)
791]
792[Add build system profiling to build system
793Ian Lynagh <igloo@earth.li>**20110123151408
794 Ignore-this: 75717810be32d60323980f9fd1baa853
795]
796[Fix ghci in stage3
797Ian Lynagh <igloo@earth.li>**20110123120232]
798[Remove use of non-existent $$(dir) variable in the rts ghc.mk
799Ian Lynagh <igloo@earth.li>**20110123021815]
800[Add some missing dependencies
801Ian Lynagh <igloo@earth.li>**20110123004208]
802[Tweak some deps to avoid multiple $(wildcard ...)s
803Ian Lynagh <igloo@earth.li>**20110123001045
804 Ignore-this: 38e53cb6f6b4f27c771ae0ed341f8958
805 Note that some things depending on the rts/includes header files now
806 depend on more files: They used to include depend on includes/*.h, but
807 now they also depend on header files in subdirectories. As far as I can
808 see this was a bug.
809]
810[Use := when assigning the result of $(wildcard ...)
811Ian Lynagh <igloo@earth.li>**20110122224532
812 Ignore-this: 67e2ca2ffbcffb5b7f55bd60c17fc6cf
813 Avoids repeated evaluations of things that need system calls etc
814]
815[Simplify the build system, and remove 2 phases
816Ian Lynagh <igloo@earth.li>**20110122190928
817 Ignore-this: 7b6184088befcbc44ea47b2f4abf85a9
818 From
819     http://hackage.haskell.org/trac/ghc/wiki/Building/Architecture/Idiom/PhaseOrdering
820 
821 Phase 0:
822     Includes: package-data.mk files for things built by the
823               bootstrapping compiler.
824     Builds:   the dependency files for hsc2hs and genprimopcode. We need
825               to do this now, as hsc2hs needs to be buildable in phase 1's
826               includes (so that we can make the hpc library's .hs source
827               files, which in turn is necessary for making its dependency
828               files), and genprimopcode needs to be buildable in phase 1's
829               includes (so that we can make the primop-*.hs-incl files,
830               which are sources for the stage1 compiler library, and thus
831               necessary for making its dependency files).
832 Phase 1:
833     Includes: dependency files for things built by the bootstrapping
834               compiler.
835     Builds:   package-data.mk files for everything else. Note that this
836               requires configuring the packages, which means telling cabal
837               which ghc to use, and thus the stage1 compiler gets built
838               during this phase.
839 Phase "":
840     Includes: dependency files for everything else.
841     Builds:   Everything else.
842]
843[Manually control more of the Cabal flags for the compiler and ghc packages
844Ian Lynagh <igloo@earth.li>**20110121230552
845 Ignore-this: 652b5f6327d246d7e2e47acbca614df2
846 For some reason the Windows HEAD builder has started thinking the ghci
847 flag should be on in stage 1. This should fix it, and generally make
848 things a little more resilient.
849]
850[Remove some hardcoded makefile settings
851Ian Lynagh <igloo@earth.li>**20110121230245
852 Ignore-this: 6b1b68aebdfbe02c15518985d2ea7559
853 Now that we used cabal to configure the ghc-bin package they are no
854 longer needed.
855]
856[tweak newArray# documentation again
857Simon Marlow <marlowsd@gmail.com>**20110119140633
858 Ignore-this: ceee33428dbad7e0f5eabfa0a2590466
859]
860[Fix OSTYPE test
861Ian Lynagh <igloo@earth.li>**20110120000308
862 Ignore-this: 8fa5d5c03297cb507a166bd85675145c
863]
864[Comments only
865simonpj@microsoft.com**20110119222247
866 Ignore-this: ea531428e9093ecedb895735ed537791
867]
868[Add OSTYPE build-system variable, and use it
869simonpj@microsoft.com**20110113155023
870 Ignore-this: c4a75f0bb27a680924e57ca7075ec116
871 
872 The use is in install.mk.in, where we need to know when
873 we're on Cygwin.
874 
875 This fixes the build on my Windows box, where I have
876 both Msys and Cygwin.
877]
878[Remove an extraneous comma that stopped ghc-cabal from building
879Ian Lynagh <igloo@earth.li>**20110119222359]
880[Move some make variables around
881Ian Lynagh <igloo@earth.li>**20110119221545
882 Ignore-this: c57c93f39d72c3baef7c5f466861dd5b
883]
884[Remove a debugging 'info'
885Ian Lynagh <igloo@earth.li>**20110119203305
886 Ignore-this: ea912ba205eaae1d2bcf0cce7c13628d
887]
888[Move the PACKAGE_MAGIC evaluation inside package-data.mk
889Ian Lynagh <igloo@earth.li>**20110119203229
890 Ignore-this: 497c4e83ae75089c24d6c794c4e2891f
891]
892[Fix libraries/index.html's haddock dependency on Windows
893Ian Lynagh <igloo@earth.li>**20110119172310]
894[Add configure phases for the stage 3 compiler
895Ian Lynagh <igloo@earth.li>**20110119130629]
896[Include kfreebsdgnu in the list of Target Platforms.
897Marco Silva <marcot@marcot.eti.br>**20110118222352
898 Ignore-this: 759482baf33903b98cd837636a3f5328
899]
900[Fix documentation bug: newArray# accepts word count, not byte count.
901Edward Z. Yang <ezyang@mit.edu>**20110118221834
902 Ignore-this: 8daab134bf72a740b89d273fb4e983d5
903]
904[Update the location of libffi.dll.a
905Ian Lynagh <igloo@earth.li>**20110118164225
906 As far as I can see this has been wrong for some time, but only bit
907 recently.
908]
909[Update the generics docs; pointed out by Christian Maeder
910Ian Lynagh <igloo@earth.li>**20110117214632]
911[ghc-cabal now adds the language flag being used
912Ian Lynagh <igloo@earth.li>**20110117184833
913 Ignore-this: 8198892ef7f8009561d3181425cde942
914 This means we get -XHaskell98 added to the list of flags, just like we
915 would if we were building with Cabal.
916]
917[Reinstate the OS X flags in the LDFLAGS etc variables
918Ian Lynagh <igloo@earth.li>**20110117200540
919 Ignore-this: 9261baa1843100f65b02fb91c1a0d225
920 I expect this will fix:
921 http://www.haskell.org/pipermail/cvs-ghc/2011-January/059098.html
922]
923[Add NondecreasingIndentation extension to ghc-bin
924Ian Lynagh <igloo@earth.li>**20110117200427
925 Ignore-this: b6b029ee6dfbda482c91d17e835f9000
926]
927[Change an "if ... else return ()" into a "when"
928Ian Lynagh <igloo@earth.li>**20110117191714
929 Ignore-this: 7de58b728e4fce7f86d7d24a3089e6c7
930]
931[Add NondecreasingIndentation to the list of extensions in ghc-pkg
932Ian Lynagh <igloo@earth.li>**20110117190610
933 Ignore-this: 20ce8144b7b64d1f67de2f6983717da3
934]
935[Add NondecreasingIndentation to the list of extensions in the ghc package
936Ian Lynagh <igloo@earth.li>**20110117190404
937 Ignore-this: 516b45e93c1b3bbb66da5414d9aabef1
938]
939[Fix deps on the ghc package
940Ian Lynagh <igloo@earth.li>**20110117173010
941 The standard libraries/$depname scheme doesn't apply, so we need to
942 handle it specially.
943]
944[Tidy up gmp cleaning
945Ian Lynagh <igloo@earth.li>**20110117121155
946 Ignore-this: 61d9a57d14b70732f62d6b2c8d6d197a
947]
948[Remove redundant libraries/cabal-bin.hs
949Ian Lynagh <igloo@earth.li>**20110116194919
950 Ignore-this: 13b4a8d26fa06ec952351603c3bb40ee
951]
952[Turn off dtrace unless you override USE_DTRACE
953Ian Lynagh <igloo@earth.li>**20110116180306
954 Ignore-this: beafc2002091fa7f0e66666004c870a5
955 There are problems with dtrace on 64bit 10.5. For now at least, we
956 just turn dtrace off unless you override USE_DTRACE
957]
958[Simplify a bit of makefile
959Ian Lynagh <igloo@earth.li>**20110116175218
960 Ignore-this: 18f02e40e36eca2e2cab79c152c72541
961]
962[Tweak Windows phase ordering
963Ian Lynagh <igloo@earth.li>**20110116173459
964 Ignore-this: bb8a70741be4574edc149349acd0f4be
965]
966[Handle dependencies of programs on libraries correctly
967Ian Lynagh <igloo@earth.li>**20110116155627]
968[It's not clear if LDFLAGS flags will be given to gcc or ld,
969Ian Lynagh <igloo@earth.li>**20110116151230
970 Ignore-this: a6a2d0b1f550c922c32f6f252e4e3285
971 and they accept different flags, so for now do nothing
972]
973[Fix cross-package dependency generation on Windows
974Ian Lynagh <igloo@earth.li>**20110116150901
975 Ignore-this: f78baaa7074ca36a6a4ff8a7e6f2e35
976]
977[Add some Windows-only CONFIGURE_PHASEs
978Ian Lynagh <igloo@earth.li>**20110116150826
979 Ignore-this: abf1bf498609107eb206b22d483613de
980]
981[Simplify, and future-proof, a dependency in the build system
982Ian Lynagh <igloo@earth.li>**20110116020035
983 Ignore-this: d089133430828d041b3601b1e9c8b22a
984]
985[Remove an unnecessary phase, and some unnecessary deps
986Ian Lynagh <igloo@earth.li>**20110116015943
987 Ignore-this: e649b072d006db5db97aee26d3753f65
988 now that cross-package deps are tracked correctly.
989]
990[We can now pass -include-pkg-deps to the bootstrapping compiler
991Ian Lynagh <igloo@earth.li>**20110116015714
992 Ignore-this: bdfed941124bb93111f117800be5f2d8
993]
994[Remove some flags that are redundant now GHC gets configured by Cabal
995Ian Lynagh <igloo@earth.li>**20110116003154
996 Ignore-this: 43a023c5103b72c91d53cf3bed7a4c50
997]
998[Change some HC_OPTS var handling
999Ian Lynagh <igloo@earth.li>**20110116003104
1000 Ignore-this: 629f4a3d37028f71a477c22ed4e8591e
1001 In particular, this means ghc gets built with -rtsopt, -threaded, etc again.
1002]
1003[Remove some unnecessary workarounds
1004Ian Lynagh <igloo@earth.li>**20110116002803
1005 Ignore-this: 5ecc62f765522c08c44aa0814c5b840e
1006 We can now rely on cross-package deps working properly, as we require
1007 GHC 6.12.
1008]
1009[Tidy up a bit
1010Ian Lynagh <igloo@earth.li>**20110116001121
1011 Ignore-this: a2baabc6da0cf2877507b7833d5b0fc7
1012]
1013[Build system improvements
1014Ian Lynagh <igloo@earth.li>**20110115231927
1015 Ignore-this: 92ea6514addc8aa8734d7e0eb61b50cb
1016 We no longer use dummy-ghc; instead we don't configure most packages
1017 until the stage1 compiler is available.
1018   
1019 We also now use Cabal for building the ghc-bin package.
1020 
1021 There are a couple more sanity checks too.
1022]
1023[Whitespace tweak
1024Ian Lynagh <igloo@earth.li>**20110115214149
1025 Ignore-this: 3e564566f311be473e94f6af609bdeaa
1026]
1027[Fix libffi build rules
1028Ian Lynagh <igloo@earth.li>**20110115202104
1029 Ignore-this: 57e1763d2079301b0165be7deba29c85
1030 Fixes a rare race when both libHSffi.a and libHSffi_p.a were being built
1031 at the same time:
1032 
1033 "cp" libffi/dist-install/build/libffi.a libffi/dist-install/build/libHSffi.a
1034 "cp" libffi/dist-install/build/libffi.a libffi/dist-install/build/libHSffi.a
1035 "cp" libffi/dist-install/build/libffi.so libffi/dist-install/build/libHSffi-ghc7.1.20110115.so
1036 cp: cannot create regular file `libffi/dist-install/build/libHSffi.a': File exists
1037]
1038[Fix Trac #4874: specialisation of INLINABLE things
1039simonpj@microsoft.com**20110114163227
1040 Ignore-this: b90543117ebddaf3bbeeaf0af0c18699
1041 
1042 Johan discovered that when INLINABLE things are specialised
1043 bad things can happen. This patch implements a hack -- but
1044 it's a simple hack and it solves the problem.
1045 
1046 See Note [Inline specialisations].
1047 
1048 The hack part is that really INLINABLE should not cause *any* loss
1049 optimisation, and it does; see Note [Don't w/w INLINABLE things] in
1050 WorkWrap.
1051]
1052[Comments only
1053simonpj@microsoft.com**20110114162959
1054 Ignore-this: f76d4d8f527c3fcd2598ec8cc5fd3049
1055]
1056[Fix a buglet in postInlineUnconditionally
1057simonpj@microsoft.com**20110114162927
1058 Ignore-this: 7a7b8610ef863907843d4ae36a8a1a3c
1059 
1060 Under obscure circumstances (actually only shown up when fixing something
1061 else) it was possible for a variable binding to be discarded although
1062 it was still used.  See Note [Top level and postInlineUnconditionally]
1063]
1064[cope with empty libraries/stamp directory (in git repo)
1065Simon Marlow <marlowsd@gmail.com>**20110114142406
1066 Ignore-this: 6e95c44368d784f86a0c1c1d1e24d810
1067]
1068[add .gitignore
1069Simon Marlow <marlowsd@gmail.com>**20110114142353
1070 Ignore-this: 23d7cabd2b04eedfe4c33ad94a120474
1071]
1072[Fix longstanding bug in C-- inlining for functions calls.
1073Edward Z. Yang <ezyang@mit.edu>**20110113130654
1074 Ignore-this: 79001003b1f3cc5005207ccfed980c21
1075]
1076[fix for remote repos without -r
1077Simon Marlow <marlowsd@gmail.com>**20110113131147
1078 Ignore-this: 3ddd8a4c616cad01a2dbdb500fb54279
1079]
1080[add a version of packages that stores all the repos in git
1081Simon Marlow <marlowsd@gmail.com>**20110113111733
1082 Ignore-this: fcca2eb2e753ee20bb5abce7f30f5205
1083]
1084[add the -r flag from darcs-all
1085Simon Marlow <marlowsd@gmail.com>**20110113111654
1086 Ignore-this: ada88377bd95ebb9c668dd48954f321e
1087]
1088[Make Template Haskell classInstances function return [ClassInstance]
1089simonpj@microsoft.com**20110113111421
1090 Ignore-this: d14381f0a94170965414dd8724188356
1091 
1092 This is a recently-introduce function, which was returning
1093 a [Name], being the names of the dfuns.  But what you really
1094 want (obviously!) is the ClassInstances, and we have a TH type
1095 for that.
1096 
1097 This is an API change, so don't merge into GHC 7.0.  But it's
1098 a new part of TH which is still settling down.
1099 
1100 Fixes Trac #4863.
1101]
1102[Improve the finder's error messages
1103simonpj@microsoft.com**20110113111233
1104 Ignore-this: ec4819b0a44af9fd03dc0a8b8e13699d
1105 
1106 I'd done all the work to add fuzzy-match suggestions, but they
1107 weren't really being used!  Here's what you get now
1108 
1109    module Foo where
1110     import Data.Lst
1111 
1112 Foo.hs:3:1:
1113     Failed to load interface for `Data.Lst'
1114     Perhaps you meant
1115       Data.List (from base)
1116       Data.List (needs flag -package haskell2010-1.0.0.0)
1117       Data.Int (needs flag -package haskell2010-1.0.0.0)
1118     Use -v to see a list of the files searched for.
1119]
1120[White space only
1121simonpj@microsoft.com**20110113093931
1122 Ignore-this: 4e46acca5241615a3283996052a634a
1123]
1124[Produce an error message, not a crash, for HsOpApp with non-var operator
1125simonpj@microsoft.com**20110112170719
1126 Ignore-this: df0f6f2e3318f9c33a714609019b0262
1127 
1128 Fixes Trac #4877.
1129]
1130[update to work with current packages file format
1131Simon Marlow <marlowsd@gmail.com>**20110112160224
1132 Ignore-this: da73498734aadbfbf0a31389a9dc44d
1133]
1134[In configure, test that GHC generates code for the correct platform (#4819)
1135Simon Marlow <marlowsd@gmail.com>**20110107163541
1136 Ignore-this: 29541d3896f9c9bcf791510edae70254
1137 Patch supplied by the bug reporter, tidied up by me.
1138 
1139 $ ./configure --with-ghc=$HOME/fp/bin/i386-unknown-linux/ghc --build=x86_64-unknown-linux
1140 checking for gfind... no
1141 checking for find... /usr/bin/find
1142 checking for sort... /usr/bin/sort
1143 checking for GHC version date... inferred 7.1.20110107
1144 checking version of ghc... 7.0.1
1145 checking build system type... x86_64-unknown-linux-gnu
1146 checking host system type... x86_64-unknown-linux-gnu
1147 checking target system type... x86_64-unknown-linux-gnu
1148 Host platform inferred as: i386-unknown-linux
1149 Target platform inferred as: i386-unknown-linux
1150 This GHC (/home/simonmar/fp/bin/i386-unknown-linux/ghc) does not generate code for the build platform
1151    GHC target platform    : i386-unknown-linux
1152    Desired build platform : x86_64-unknown-linux
1153]
1154[Major refactoring of the type inference engine
1155simonpj@microsoft.com**20110112145604
1156 Ignore-this: 6a7fc90c9b798e89505606726cc8090e
1157 
1158 This patch embodies many, many changes to the contraint solver, which
1159 make it simpler, more robust, and more beautiful.  But it has taken
1160 me ages to get right. The forcing issue was some obscure programs
1161 involving recursive dictionaries, but these eventually led to a
1162 massive refactoring sweep.
1163 
1164 Main changes are:
1165  * No more "frozen errors" in the monad.  Instead "insoluble
1166    constraints" are now part of the WantedConstraints type.
1167 
1168  * The WantedConstraint type is a product of bags, instead of (as
1169    before) a bag of sums.  This eliminates a good deal of tagging and
1170    untagging.
1171 
1172  * This same WantedConstraints data type is used
1173      - As the way that constraints are gathered
1174      - As a field of an implication constraint
1175      - As both argument and result of solveWanted
1176      - As the argument to reportUnsolved
1177 
1178  * We do not generate any evidence for Derived constraints. They are
1179    purely there to allow "impovement" by unifying unification
1180    variables.
1181 
1182  * In consequence, nothing is ever *rewritten* by a Derived
1183    constraint.  This removes, by construction, all the horrible
1184    potential recursive-dictionary loops that were making us tear our
1185    hair out.  No more isGoodRecEv search either. Hurrah!
1186 
1187  * We add the superclass Derived constraints during canonicalisation,
1188    after checking for duplicates.  So fewer superclass constraints
1189    are generated than before.
1190 
1191  * Skolem tc-tyvars no longer carry SkolemInfo.  Instead, the
1192    SkolemInfo lives in the GivenLoc of the Implication, where it
1193    can be tidied, zonked, and substituted nicely.  This alone is
1194    a major improvement.
1195 
1196  * Tidying is improved, so that we tend to get t1, t2, t3, rather
1197    than t1, t11, t111, etc
1198 
1199    Moreover, unification variables are always printed with a digit
1200    (thus a0, a1, etc), so that plain 'a' is available for a skolem
1201    arising from a type signature etc. In this way,
1202      (a) We quietly say which variables are unification variables,
1203          for those who know and care
1204      (b) Types tend to get printed as the user expects.  If he writes
1205              f :: a -> a
1206              f = ...blah...
1207          then types involving 'a' get printed with 'a', rather than
1208          some tidied variant.
1209 
1210  * There are significant improvements in error messages, notably
1211    in the "Cannot deduce X from Y" messages.
1212]
1213[Fix installation on cygwin
1214Ian Lynagh <igloo@earth.li>**20110111194838
1215 Ignore-this: fe923d0619da3bd3a34968106c92fdab
1216]
1217[Do dependency analysis when kind-checking type declarations
1218simonpj@microsoft.com**20110110110351
1219 Ignore-this: 17a8dee32694d3e1835cf7bb02d3abb5
1220 
1221 This patch fixes Trac #4875.  The main point is to do dependency
1222 analysis on type and class declarations, and kind-check them in
1223 dependency order, so as to improve error messages.
1224 
1225 This patch means that a few programs that would typecheck before won't
1226 typecheck any more; but before we were (naughtily) going beyond
1227 Haskell 98 without any language-extension flags, and Trac #4875
1228 convinces me that doing so is a Bad Idea.
1229 
1230 Here's an example that won't typecheck any more
1231        data T a b = MkT (a b)
1232        type F k = T k Maybe
1233 
1234 If you look at T on its own you'd default 'a' to kind *->*;
1235 and then kind-checking would fail on F.
1236 
1237 But GHC currently accepts this program beause it looks at
1238 the *occurrences* of T.
1239]
1240[Move imports around (no change in behaviour)
1241simonpj@microsoft.com**20110110105647
1242 Ignore-this: d618cabbc52be7d7968de1e0bdd44082
1243]
1244[Make fuzzy matching a little less eager for short identifiers
1245simonpj@microsoft.com**20110107102855
1246 Ignore-this: a753643e88433d74b44a480cc0f4170c
1247 
1248 For single-character identifiers we now don't make any suggestions
1249 See comments in Util.fuzzyLookup
1250]
1251[Fix Trac #4870: get the inlining for an imported INLINABLE Id
1252simonpj@microsoft.com**20110105002712
1253 Ignore-this: 60c0192eb48590c2e6868d15ba8f84ce
1254 
1255 We need the unfolding even for a *recursive* function (indeed
1256 that's the point) and I was using the wrong function to get it
1257 (idUnfolding rather than realIdUnfolding).
1258]
1259[Rejig the includes/ installation rules
1260Ian Lynagh <igloo@earth.li>**20110109181158
1261 They're a little nicer now, and a regression in the cygwin build is
1262 fixed (the $i in the destination wasn't surviving being passed through
1263 cygpath).
1264]
1265[Make DESTDIR an absolute path when installing; fixes #4883
1266Ian Lynagh <igloo@earth.li>**20110108171635]
1267[Add utils/ghc-cabal/Makefile
1268Ian Lynagh <igloo@earth.li>**20110108144049]
1269[Remove redundant import
1270Ian Lynagh <igloo@earth.li>**20110108130047
1271 Ignore-this: 1c7fdec77b48319c845c9593b5fb94af
1272]
1273[Improve error message of :set in ghci (ticket #4190).
1274Michal Terepeta <michal.terepeta@gmail.com>**20101130211505
1275 Ignore-this: ccc8a0816a900ba8c4a966285a465b23
1276]
1277[Improve error message when importing data constructors (ticket #4058).
1278Michal Terepeta <michal.terepeta@gmail.com>**20101127211338
1279 Ignore-this: 3289a08f0391dd90dfef2e0403a04ccd
1280]
1281[catch SIGTSTP and save/restore terminal settings (#4460)
1282Simon Marlow <marlowsd@gmail.com>**20110107124042
1283 Ignore-this: 38f7f27bf75178899f466404c048241d
1284 As far as I can tell, it is the responsibility of the program to save
1285 and restore its own terminal settings across a suspend/foreground, the
1286 shell doesn't do it (which seems odd).  So I've added a signal handler
1287 for SIGTSTP to the RTS which will save and restore the terminal
1288 settings iff we modified them with hSetBuffering or hSetEcho (we
1289 already restore them at exit time in these cases).
1290]
1291[comment updates
1292Simon Marlow <marlowsd@gmail.com>**20110107094236
1293 Ignore-this: c2b30b0c98645e2847a2749c7fdc167f
1294]
1295[On Cygwin, use a Cygwin-style path for /bin/install's destination
1296Ian Lynagh <igloo@earth.li>**20110106223030
1297 
1298 cygwin's /bin/install doesn't set file modes correctly if the
1299 destination path is a C: style path:
1300 
1301 $ /bin/install -c -m 644 foo /cygdrive/c/cygwin/home/ian/foo2
1302 $ /bin/install -c -m 644 foo c:/cygwin/home/ian/foo3
1303 $ ls -l foo*
1304 -rw-r--r-- 1 ian None 0 2011-01-06 18:28 foo
1305 -rw-r--r-- 1 ian None 0 2011-01-06 18:29 foo2
1306 -rwxrwxrwx 1 ian None 0 2011-01-06 18:29 foo3
1307 
1308 This causes problems for bindisttest/checkBinaries.sh which then
1309 thinks that e.g. the userguide HTML files are binaries.
1310 
1311 We therefore use a /cygdrive path if we are on cygwin
1312]
1313[Fix mkUserGuidePart program name on Windows
1314Ian Lynagh <igloo@earth.li>**20110106143707]
1315[add comment to remind people to update driver/gcc/gcc.c
1316Simon Marlow <marlowsd@gmail.com>**20110106152402
1317 Ignore-this: c07d7ac11eb9221ef821f78aab1807cb
1318]
1319[use Win32 CreateProcess() rather than mingw spawnv() (#4531)
1320Simon Marlow <marlowsd@gmail.com>**20110106133834
1321 Ignore-this: 4c0947853549dad034622c044391af6c
1322]
1323[update paths now that we upgraded gcc to 4.5.0
1324Simon Marlow <marlowsd@gmail.com>**20110106133729
1325 Ignore-this: f8f9bcad984fdd472e0ae958b66bea9d
1326]
1327[fix markup
1328Simon Marlow <marlowsd@gmail.com>**20110106093152
1329 Ignore-this: 555b6e39ae6b5a177b03c5edffc169ab
1330]
1331[fix up multi-line GHCi patch (#4316)
1332Simon Marlow <marlowsd@gmail.com>**20110105154548
1333 Ignore-this: 53d5d489bd2a792c01f2cc56a11f3ce6
1334]
1335[multiline commands in GHCi #4316
1336Vivian McPhail <haskell.vivian.mcphail@gmail.com>**20101105051308
1337 This patch adds support for multiline commands in GHCi.
1338 
1339 The first line of input is lexed.  If there is an active
1340 layout context once the lexer reaches the end of file, the
1341 user is prompted for more input.
1342 
1343 Multiline input is exited by an empty line and can be escaped
1344 with a user interrupt.
1345 
1346 Multiline mode is toggled with `:set +m`
1347]
1348[Replace a #if with a Haskell conditional
1349Ian Lynagh <igloo@earth.li>**20110105183011
1350 Ignore-this: f08f3a4356586efab2725ad8704b2eba
1351]
1352[Whitespace only in X86.Ppr
1353Ian Lynagh <igloo@earth.li>**20110105171124]
1354[Fix error compiling AsmCodeGen.lhs for PPC Mac (unused makeFar addr)
1355naur@post11.tele.dk**20101219213555
1356 Ignore-this: ab25d5f2e2ebe163547d5babaf4b1dbf
1357]
1358[Define cTargetArch and start to use it rather than ifdefs
1359Ian Lynagh <igloo@earth.li>**20110104220013
1360 Using Haskell conditionals means the compiler sees all the code, so
1361 there should be less rot of code specific to uncommon arches. Code
1362 for other platforms should still be optimised away, although if we want
1363 to support targetting other arches then we'll need to compile it
1364 for-real anyway.
1365]
1366[Fix error compiling AsmCodeGen.lhs for PPC Mac (rtsPackageId)
1367naur@post11.tele.dk**20101219212530
1368 Ignore-this: 946f6d3e0d3c3ddf2dc07b85e1f82d85
1369]
1370[Rename the c*Platform variables to c*PlatformString
1371Ian Lynagh <igloo@earth.li>**20110104210250]
1372[Fix #4829 (build does not respect --with-gcc option)
1373gwright@antiope.com**20101221133233
1374 Ignore-this: 37918feb82f911c2beb75915b6e8b97b
1375 
1376 This patch fixes what seems to be the last problem with the --with-gcc
1377 option.  On OS X, we need to pass the path to gcc to dtrace as the
1378 preprocessor.  (Internally, dtrace on OS X sets the default preprocessor
1379 to /usr/bin/gcc.)  ATM, dtrace is only supported on OS X, so we don't
1380 need any conditionalization.  If dtrace is ported to other platforms,
1381 we might need to change this. However, usage on other platforms will
1382 probably be similar to OS X, since many of Apple's changes are to
1383 use the gnu toolchain instead of the Sun toolchain.
1384   
1385]
1386[Drop a seven years old workaround for happy
1387Matthias Kilian <kili@outback.escape.de>**20101231192343
1388 Ignore-this: a9348c91292c113bd967464fbe859f1f
1389]
1390[Add gcc and ld flags to --info output
1391Ian Lynagh <igloo@earth.li>**20101220173520]
1392[Fix Trac #4525: report type errors in terms of the immediate type synonym
1393simonpj@microsoft.com**20101224082520
1394 Ignore-this: a3bd076bfe0e1c6f575b106f77f326c6
1395 
1396 This small change means that if you have
1397      type Age = Int
1398 and you try to unify Age and Bool, you'll get a complaint about
1399 not matching Age and Bool, rather than Int and Bool.  See the notes
1400 with Trac #4525
1401]
1402[Comments only
1403simonpj@microsoft.com**20101224082310
1404 Ignore-this: 1f69fa3244663b653607093efcdf7b0
1405]
1406[Implement fuzzy matching for the Finder
1407simonpj@microsoft.com**20101222175400
1408 Ignore-this: 4dfbbc07bcb59c5f4cee9a902c89d63e
1409 
1410 ..so that you get a more helpful message when
1411 you mis-spell a module name in an 'import'.
1412 
1413 Validates, but not fully tested.
1414 
1415 Based on Max's patch in Trac #2442, but heavily refactored.
1416]
1417[Implement fuzzy matching for the renamer
1418simonpj@microsoft.com**20101222175306
1419 Ignore-this: 66478736249de793a61612f184d484b0
1420 
1421 ...so that you get helpful suggestions when you mis-spell a name
1422 Based on Max's patch in Trac #2442, but heavily refactored.
1423]
1424[Add fuzzyLookup, a variant of fuzzyMatch
1425simonpj@microsoft.com**20101222175124
1426 Ignore-this: f0eafaf275b9edffee176f2fb4effe2f
1427 
1428 Plus, I changed quite a bit of layout to make the lines shorter.
1429]
1430[White space only
1431simonpj@microsoft.com**20101222175001
1432 Ignore-this: ddabada2042f4529e83d1c1ecb052306
1433]
1434[Layout and white space only
1435simonpj@microsoft.com**20101222174950
1436 Ignore-this: bf4e4fd9d39714d0461ab799d6b8ed91
1437]
1438[Tidy up rebindable syntax for MDo
1439simonpj@microsoft.com**20101222132210
1440 Ignore-this: b40ae8709e5a39d75f2b2813169af215
1441 
1442 For a long time an 'mdo' expression has had a SyntaxTable
1443 attached to it.  However, we're busy deprecating SyntaxTables
1444 in favour of rebindable syntax attached to individual Stmts,
1445 and MDoExpr was totally inconsistent with DoExpr in this
1446 regard.
1447 
1448 This patch tidies it all up.  Now there's no SyntaxTable on
1449 MDoExpr, and 'modo' is generally handled much more like 'do'.
1450 
1451 There is resulting small change in behaviour: now MonadFix is
1452 required only if you actually *use* recursion in mdo. This
1453 seems consistent with the implicit dependency analysis that
1454 is done for mdo.
1455 
1456 Still to do:
1457   * Deal with #4148 (this patch is on the way)
1458   * Get rid of the last remaining SyntaxTable on HsCmdTop
1459]
1460[Make the occurrence analyser track preInlineUnconditionally
1461simonpj@microsoft.com**20101222131156
1462 Ignore-this: 82edb06bcca6106327c2cce9d78c4e61
1463 
1464 This fixes a somewhat obscure situation in which an
1465 over-optimistic use of "occurs once" led to an infinite
1466 sequence of simplifier iterations.  Se Note [Cascading inlines]
1467 for the details.
1468 
1469 This showed up when compiling rather large DPH programs, which
1470 run lots of iterations of the simplifier, which in turn made
1471 compilation take much longer than necessary.
1472]
1473[Make mkDFunUnfolding more robust
1474simonpj@microsoft.com**20101222130854
1475 Ignore-this: 10bb4168a7080c843f6613043354151b
1476 
1477 It now uses tcSplitDFunTy, which is designed for the purpose and
1478 allows arbitrary argument types to the dfun, rather than
1479 tcSplitSigmaTy.  This generality is used in DPH, which has
1480 internally-generated dfuns with impliciation-typed arguments.
1481 
1482 To do this I had to make tcSplitDFunTy return the number of
1483 arguments, so there are some minor knock-on effects in other
1484 modules.
1485]
1486[Count allocations more accurately
1487Simon Marlow <marlowsd@gmail.com>**20101221152956
1488 Ignore-this: 33a4ed3a77bf35f232aa5c9078e8e380
1489 The allocation stats (+RTS -s etc.) used to count the slop at the end
1490 of each nursery block (except the last) as allocated space, now we
1491 count the allocated words accurately.  This should make allocation
1492 figures more predictable, too.
1493 
1494 This has the side effect of reducing the apparent allocations by a
1495 small amount (~1%), so remember to take this into account when looking
1496 at nofib results.
1497]
1498[Add a simple arity analyser
1499simonpj@microsoft.com**20101221165800
1500 Ignore-this: d5f3a9f56404d61bb7f374c875b42c49
1501 
1502 I've wanted to do this for ages, but never gotten around to
1503 it.  The main notes are in Note [Arity analysis] in SimplUtils.
1504 
1505 The motivating example for arity analysis is this:
1506 
1507   f = \x. let g = f (x+1)
1508           in \y. ...g...
1509 
1510 What arity does f have?  Really it should have arity 2, but a naive
1511 look at the RHS won't see that.  You need a fixpoint analysis which
1512 says it has arity "infinity" the first time round.
1513 
1514 This makes things more robust to the way in which you write code.  For
1515 example, see Trac #4474 which is fixed by this change.
1516 
1517 Not a huge difference, but worth while:
1518 
1519         Program           Size    Allocs   Runtime   Elapsed
1520 --------------------------------------------------------------------------------
1521             Min          -0.4%     -2.2%    -10.0%    -10.0%
1522             Max          +2.7%     +0.3%     +7.1%     +6.9%
1523  Geometric Mean          -0.3%     -0.2%     -2.1%     -2.2%
1524 
1525 I don't really believe the runtime numbers, because the machine was
1526 busy, but the bottom line is that not much changes, and what does
1527 change reliably (allocation and size) is in the right direction.
1528]
1529[Miscellaneous tidying up and refactoring
1530simonpj@microsoft.com**20101221161931
1531 Ignore-this: 7706d3065e6fc1defafe1cb8975b9969
1532]
1533[Comments only
1534simonpj@microsoft.com**20101221161918
1535 Ignore-this: 3e269a62da5cbec72d3e4b8328689628
1536]
1537[Single-method classes are implemented with a newtype
1538simonpj@microsoft.com**20101221161911
1539 Ignore-this: 4ca00f0b367fbeb8146146bc53116eb7
1540 
1541 This patch changes things so that such classes rely on the coercion
1542 mechanism for inlining (since the constructor is really just a cast)
1543 rather than on the dfun mechanism, therby removing some needless
1544 runtime indirections.
1545]
1546[For single-method classes use newtypes
1547simonpj@microsoft.com**20101101080736
1548 Ignore-this: d3851f92eb2385501411da57066b775e
1549 
1550 This clears up an awkward hack for exprIsConApp_maybe, and
1551 works better too.  See Note [Single-method classes] in
1552 TcInstDcls.
1553]
1554[boundTaskExiting: don't set task->stopped unless this is the last call (#4850)
1555Simon Marlow <marlowsd@gmail.com>**20101221115807
1556 Ignore-this: 7e1b990aa08b3ea9cdaa9385d8e41e48
1557 The bug in this case was that we had a worker thread making a foreign
1558 call which invoked a callback (in this case it was performGC, I
1559 think).  When the callback ended, boundTaskExiting() was setting
1560 task->stopped, but the Task is now per-OS-thread, so it is shared by
1561 the worker that made the original foreign call.  When the foreign call
1562 returned, because task->stopped was set, the worker was not placed on
1563 the queue of spare workers.  Somehow the worker woke up again, and
1564 found the spare_workers queue empty, which lead to a crash.
1565 
1566 Two bugs here: task->stopped should not have been set by
1567 boundTaskExiting (this broke when I split the Task and InCall structs,
1568 in 6.12.2), and releaseCapabilityAndQueueWorker() should not be
1569 testing task->stopped anyway, because it should only ever be called
1570 when task->stopped is false (this is now an assertion).
1571]
1572[releaseCapabilityAndQueueWorker: task->stopped should be false (#4850)
1573Simon Marlow <marlowsd@gmail.com>**20101221114911
1574 Ignore-this: b9c430a4bc9d2e0c7f4140d6d6971eae
1575]
1576[Fix Windows build
1577Simon Marlow <marlowsd@gmail.com>**20101221102101
1578 Ignore-this: f4773e06d030a335c9ac721af193b8d2
1579]
1580[raiseExceptionHelper: update tso->stackobj->sp before calling threadStackOverflow (#4845)
1581Simon Marlow <marlowsd@gmail.com>**20101221101411
1582 Ignore-this: 48495131fcc8c548882a470c2509f9f5
1583]
1584[add 'make re2' for rebuilding stage2 (similarly re1 and re3)
1585Simon Marlow <marlowsd@gmail.com>**20101221100254
1586 Ignore-this: 5c0afe3810b66a5b6e53a3a0fe933945
1587]
1588[fix warning
1589Simon Marlow <marlowsd@gmail.com>**20101216160415
1590 Ignore-this: 54a0eedfa5b7fc15c31dffffb1b10aad
1591]
1592[Small improvement to CorePrep
1593simonpj@microsoft.com**20101220123715
1594 Ignore-this: d0490225ed1895a1a5b97d786ed44260
1595 
1596 This change avoids unnecessary bindings. Example
1597 
1598      foo (let fn = \x.blah in
1599           in fn)
1600 
1601 We were generating something stupid like
1602 
1603     let fn = \x.blah in
1604     let fn' = \eta. fn eta
1605     in foo fn
1606 
1607 Now we don't.  The change is quite small.
1608 
1609 Thanks to Ben for showing me an example of this happening.
1610]
1611[Fix warnings
1612Ian Lynagh <igloo@earth.li>**20101219202711
1613 Ignore-this: 898015b086f684de5371bf97a23b9e2e
1614]
1615[Small refactoring
1616Ian Lynagh <igloo@earth.li>**20101219194032]
1617[Drop GhcWithLlvmCodeGen configuration bits
1618Matthias Kilian <kili@outback.escape.de>**20101219180239
1619 Ignore-this: 815ed46be7650792f85807c232edfcc
1620 The LLVM code generator is always built unconditionally, so both the
1621 configuration variable in mk/config.mk.in as well as the string in
1622 compilerInfo can be removed.
1623]
1624[Pass --hoogle to haddock; fixes trac #4521
1625Ian Lynagh <igloo@earth.li>**20101219125243]
1626[vectoriser: don't always pass superclass dictionaries to PA dfuns
1627Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101218234838
1628 Ignore-this: 77c71976db8fc63aeb83f4abdba994d8
1629 
1630 This is just a guess at how this should work.
1631]
1632[vectoriser: delete dead code
1633Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101218125350
1634 Ignore-this: 437eea71ad15ad5dc7902e596597c577
1635]
1636[vectoriser: adapt to new superclass story part I (dictionary construction)
1637Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101218114953
1638 Ignore-this: 29c9aa46a1622beaae1dcefc4c482a30
1639]
1640[Replace uses of the old try function with the new one
1641Ian Lynagh <igloo@earth.li>**20101218230827
1642 Ignore-this: 5dd6c1a4142405aa1aab3fc4ec07eea6
1643]
1644[Replace uses of the old catch function with the new one
1645Ian Lynagh <igloo@earth.li>**20101218213350]
1646[Create ~/.ghc/ if it doesn't already exist; fixes trac #4522
1647Ian Lynagh <igloo@earth.li>**20101218184925]
1648[Document GADTSyntax extension
1649Ian Lynagh <igloo@earth.li>**20101218150121]
1650[Implement GADTSyntax extension
1651Ian Lynagh <igloo@earth.li>**20101218144550]
1652[Whitespace-only in rts/Linker.c
1653Ian Lynagh <igloo@earth.li>**20101217234124]
1654[Add some casts to fix warnings; patch from Greg Wright
1655Ian Lynagh <igloo@earth.li>**20101217223811]
1656[Put an up-to-date Makefile in docs/Makefile
1657Ian Lynagh <igloo@earth.li>**20101217223707
1658 It doesn't do anything useful yet, but it works with the new build system
1659]
1660[do not compile part of shared lib RTS with -fno-PIC on Solaris
1661Karel Gardas <karel.gardas@centrum.cz>**20101217085133
1662 Ignore-this: 8c8dbb45cac0578a58a3557f1e03c66
1663]
1664[provide shared libraries support on i386-unknown-solaris2 platform
1665Karel Gardas <karel.gardas@centrum.cz>**20101217084617
1666 Ignore-this: b6079c6a39a71200a1ee863573e40828
1667]
1668[fix CPP detection of Solaris in NCG
1669Karel Gardas <karel.gardas@centrum.cz>**20101217084510
1670 Ignore-this: 9d1ce59d469294eab1f0cbc697e48d69
1671]
1672[Fix checkBinaries on OS X
1673Ian Lynagh <igloo@earth.li>**20101216201121]
1674[Remove redundant HpcMap and HpcSet wrappers around Data.{Map,Set}
1675Ian Lynagh <igloo@earth.li>**20101216190605]
1676[Use "-perm -u+x" rather than "-executable" to find executables
1677Ian Lynagh <igloo@earth.li>**20101216145235
1678 On Windows, -executable is matching the html docs.
1679]
1680[Remove a debugging print
1681Ian Lynagh <igloo@earth.li>**20101216011459]
1682[__GLASGOW_HASKELL__ >= 604 is now always true
1683Ian Lynagh <igloo@earth.li>**20101215214656]
1684[Remove more dead code now we require GHC >= 6.12
1685Ian Lynagh <igloo@earth.li>**20101215213715]
1686[refactor and tidy up the section on RTS options
1687Simon Marlow <marlowsd@gmail.com>**20101216123151
1688 Ignore-this: 9cdafd687351d8a3ff879b64347f85d3
1689]
1690[Related to #4826: Some minor tweaks to the wording of the User Guide, section 4.16
1691Orphi <MathematicalOrchid@hotmail.com>**20101209170440
1692 Ignore-this: c3d942d58594be7d4c2eb4dc3a22f19
1693]
1694[FIX #4826 partial: Add -rtsopts and -with-rtsopts to User Guide section 4.11.6
1695Orphi <MathematicalOrchid@hotmail.com>**20101209165152
1696 Ignore-this: 2fc1c0abbb783695773ab0f9c013bbaa
1697]
1698[FIX #4826 partially: Change -f to -? in User Guide section F4.16
1699Orphi <MathematicalOrchid@hotmail.com>**20101209144148
1700 Ignore-this: 73410b350e80c8943ae722dec8dea44b
1701]
1702[fix #3910
1703Simon Marlow <marlowsd@gmail.com>**20101216114452
1704 Ignore-this: 410e95e188344a523520e192a3fb58ea
1705]
1706[remove an optimisation that wasn't
1707Simon Marlow <marlowsd@gmail.com>**20101215152656
1708 Ignore-this: e8413f58e8292c6e7463087d885b3a7d
1709]
1710[fix a warning
1711Simon Marlow <marlowsd@gmail.com>**20101216105723
1712 Ignore-this: ed6024378021a698ce638267ed3e21ab
1713]
1714[use EXTERN_INLINE instead of STATIC_INLINE to avoid some gcc warnings
1715Simon Marlow <marlowsd@gmail.com>**20101216105709
1716 Ignore-this: d4e1586cf318883a8e611b55df7fbf10
1717]
1718[remove dead code
1719Simon Marlow <marlowsd@gmail.com>**20101216104944
1720 Ignore-this: 97a04a3e37c1b28abc222a28bab3d17d
1721]
1722[fix retainer profiling: add missing case for TSO
1723Simon Marlow <marlowsd@gmail.com>**20101216103900
1724 Ignore-this: 11bda81ac159f638d719c1f6177702fb
1725]
1726[add a missing STACK case
1727Simon Marlow <marlowsd@gmail.com>**20101216102100
1728 Ignore-this: ac1c036b5cbf4209b1d10b6ab1c83f27
1729]
1730[Remove code that is dead now that we need >= 6.12 to build
1731Ian Lynagh <igloo@earth.li>**20101215201006]
1732[fix for large stack allocations
1733Simon Marlow <marlowsd@gmail.com>**20101215152419
1734 Ignore-this: d9aca17d68bd99214c126989a2318e79
1735]
1736[Implement stack chunks and separate TSO/STACK objects
1737Simon Marlow <marlowsd@gmail.com>**20101215120843
1738 Ignore-this: 73fa9460314d4a4e54456af12bef7960
1739 
1740 This patch makes two changes to the way stacks are managed:
1741 
1742 1. The stack is now stored in a separate object from the TSO.
1743 
1744 This means that it is easier to replace the stack object for a thread
1745 when the stack overflows or underflows; we don't have to leave behind
1746 the old TSO as an indirection any more.  Consequently, we can remove
1747 ThreadRelocated and deRefTSO(), which were a pain.
1748 
1749 This is obviously the right thing, but the last time I tried to do it
1750 it made performance worse.  This time I seem to have cracked it.
1751 
1752 2. Stacks are now represented as a chain of chunks, rather than
1753    a single monolithic object.
1754 
1755 The big advantage here is that individual chunks are marked clean or
1756 dirty according to whether they contain pointers to the young
1757 generation, and the GC can avoid traversing clean stack chunks during
1758 a young-generation collection.  This means that programs with deep
1759 stacks will see a big saving in GC overhead when using the default GC
1760 settings.
1761 
1762 A secondary advantage is that there is much less copying involved as
1763 the stack grows.  Programs that quickly grow a deep stack will see big
1764 improvements.
1765 
1766 In some ways the implementation is simpler, as nothing special needs
1767 to be done to reclaim stack as the stack shrinks (the GC just recovers
1768 the dead stack chunks).  On the other hand, we have to manage stack
1769 underflow between chunks, so there's a new stack frame
1770 (UNDERFLOW_FRAME), and we now have separate TSO and STACK objects.
1771 The total amount of code is probably about the same as before.
1772 
1773 There are new RTS flags:
1774 
1775    -ki<size> Sets the initial thread stack size (default 1k)  Egs: -ki4k -ki2m
1776    -kc<size> Sets the stack chunk size (default 32k)
1777    -kb<size> Sets the stack chunk buffer size (default 1k)
1778 
1779 -ki was previously called just -k, and the old name is still accepted
1780 for backwards compatibility.  These new options are documented.
1781]
1782[comments on SRC_HC_OPTS (#4829)
1783Simon Marlow <marlowsd@gmail.com>**20101214101340
1784 Ignore-this: e2bdec00f07b68e82837e77a4faf6514
1785]
1786[fix another sanity error, and refactor/tidy up
1787Simon Marlow <marlowsd@gmail.com>**20101209163919
1788 Ignore-this: d5ce953ac78e90fc0e22cd9848d26e2e
1789]
1790[Fix a bug in functorLikeTraverse, which was giving wrong answer for tuples
1791simonpj@microsoft.com**20101215123725
1792 Ignore-this: 560220e92429b5b1a6197a62f94a4ff2
1793 
1794 This bug led to Trac #4816, which is hereby fixed
1795]
1796[Improve printing for -ddump-deriv
1797simonpj@microsoft.com**20101215121955
1798 Ignore-this: 3181c948c4c2471bd99b32c5ee487a1e
1799]
1800[Tighten up what it means to be an "enumeration data constructor"
1801simonpj@microsoft.com**20101215121927
1802 Ignore-this: 459b3f9f7994a13094ed87b0768b33a8
1803 
1804 See Note [Enumeration types] in TyCon, and comments in Trac #4528
1805]
1806[Allow enumerations to have phantom arguments.
1807simonpj@microsoft.com**20101215121817
1808 Ignore-this: 32ef8cb869e6e38c2e43b3ae87b1b9a8
1809 
1810 The bytecode generator was being too eager.
1811 Fixes Trac #4528, or rather, a near variant.
1812]
1813[Instance declaration overlap allowed if *either* has -XOverlappingInstances
1814simonpj@microsoft.com**20101214180500
1815 Ignore-this: f1b1492541a7e0464ebc6adb45510a2e
1816 
1817 This satisfies Trac #3877.  Documentation is changed too.
1818 I'm not sure if this should go in 7.0.2.
1819]
1820[Fix Trac #4841: behave right with TypeSynonymInstances and NoFlexibleInstances
1821simonpj@microsoft.com**20101214174755
1822 Ignore-this: dccd707fdca84904b7885170a296ecb6
1823 
1824 When we have TypeSynonymInstances without FlexibleInstances we should still
1825 insist on a H98-style instance head, after looking through the synonym.
1826 
1827 This patch also make FlexibleInstances imply TypeSynonymInstances.  Anything
1828 else is a bit awkward, and not very useful.
1829 
1830]
1831[Fix Trac #3731: more superclass subtlety (sigh)
1832simonpj@microsoft.com**20101214180344
1833 Ignore-this: f4168e59f3164303ba7be022ba19c37b
1834 
1835 I will add more comments, but I want to commit this tonight,
1836 so the overnight builds get it.
1837]
1838[Less verbose debug print
1839simonpj@microsoft.com**20101214180248
1840 Ignore-this: e405e8545763e913155abe43daf7e36c
1841]
1842[Wibble to InstEnv.instanceHead
1843simonpj@microsoft.com**20101214082939
1844 Ignore-this: 851db517f8638a0aeb7ad461298f7e9f
1845 
1846 Fixes an accidental glitch in T1835
1847]
1848[Remove dead code now that we require the bootstrapping compiler be >= 6.12
1849Ian Lynagh <igloo@earth.li>**20101214011011]
1850[GHC 6.12 is now needed to build the HEAD
1851Ian Lynagh <igloo@earth.li>**20101214010923]
1852[Add libstdc++-4.5.0-1-mingw32-dll-6.tar.lzma to mingw tarballs
1853Ian Lynagh <igloo@earth.li>**20101213223153]
1854[Fix recursive superclasses (again).  Fixes Trac #4809.
1855simonpj@microsoft.com**20101213171511
1856 Ignore-this: b91651397918fd8f0183812f9a070073
1857 
1858 This patch finally deals with the super-delicate question of
1859 superclases in possibly-recursive dictionaries.  The key idea
1860 is the DFun Superclass Invariant (see TcInstDcls):
1861 
1862      In the body of a DFun, every superclass argument to the
1863      returned dictionary is
1864        either   * one of the arguments of the DFun,
1865        or       * constant, bound at top level
1866 
1867 To establish the invariant, we add new "silent" superclass
1868 argument(s) to each dfun, so that the dfun does not do superclass
1869 selection internally.  There's a bit of hoo-ha to make sure that
1870 we don't print those silent arguments in error messages; a knock
1871 on effect was a change in interface-file format.
1872 
1873 A second change is that instead of the complex and fragile
1874 "self dictionary binding" in TcInstDcls and TcClassDcl,
1875 using the same mechanism for existential pattern bindings.
1876 See Note [Subtle interaction of recursion and overlap] in TcInstDcls
1877 and Note [Binding when looking up instances] in InstEnv.
1878 
1879 Main notes are here:
1880 
1881   * Note [Silent Superclass Arguments] in TcInstDcls,
1882     including the DFun Superclass Invariant
1883 
1884 Main code changes are:
1885 
1886   * The code for MkId.mkDictFunId and mkDictFunTy
1887 
1888   * DFunUnfoldings get a little more complicated;
1889     their arguments are a new type DFunArg (in CoreSyn)
1890 
1891   * No "self" argument in tcInstanceMethod
1892   * No special tcSimplifySuperClasss
1893   * No "dependents" argument to EvDFunApp
1894 
1895 IMPORTANT
1896    It turns out that it's quite tricky to generate the right
1897    DFunUnfolding for a specialised dfun, when you use SPECIALISE
1898    INSTANCE.  For now I've just commented it out (in DsBinds) but
1899    that'll lose some optimisation, and I need to get back to
1900    this.
1901]
1902[Doing the smart canonicalization only if we are not simplifying a Rule LHS.
1903dimitris@microsoft.com**20101210132221
1904 Also, same thing now applies for adding superclasses.
1905 
1906]
1907[Moved canonicalisation inside solveInteract
1908dimitris@microsoft.com**20101209141215
1909 
1910 Moreover canonicalisation now is "clever", i.e. it never canonicalizes a class
1911 constraint if it can already discharge it from some other inert or previously
1912 encountered constraints. See Note [Avoiding the superclass explosion]
1913 
1914]
1915[GHCi linker: Assume non-Haskell libraries are dynamic libs
1916Ian Lynagh <igloo@earth.li>**20101213124930
1917 Ignore-this: aa153a8f6e309c7b3dae7e46bb7a9583
1918 This works around a segfault we get when trying to load libiconv.a on
1919 some platforms.
1920]
1921[Add --version support to ghc-cabal
1922Ian Lynagh <igloo@earth.li>**20101212213600
1923 Ignore-this: ef696dcb1b96a23765f9f18e75a56f5
1924]
1925[Don't link the GHC RTS into our C-only programs
1926Ian Lynagh <igloo@earth.li>**20101210185402
1927 Ignore-this: 56f620f7eb16a03e7497a161bc48458e
1928]
1929[Build a copy of ghc-cabal with the in-tree compiler, for the bindist
1930Ian Lynagh <igloo@earth.li>**20101210181123]
1931[Add a test that all programs in the bindist were built with the right GHC
1932Ian Lynagh <igloo@earth.li>**20101210161218
1933 They should use the GHC from the build tree, not the bootstrapping compiler.
1934]
1935[Fix Trac #4534: renamer bug
1936simonpj@microsoft.com**20101210084530
1937 Ignore-this: 8163bfa3a56344cfe89ad17c62e9655d
1938   
1939 The renamer wasn't attaching the right used-variables to a
1940 TransformStmt constructor.
1941 
1942 The real modification is in RnExpr; the rest is just
1943 pretty-printing and white space.
1944]
1945[White space only
1946simonpj@microsoft.com**20101210084255
1947 Ignore-this: 3fcf8a4fc8c15052c379a135951d53ea
1948]
1949[Comments only
1950simonpj@microsoft.com**20101210084116
1951 Ignore-this: 55bb1de129b1c9513751885eaa84b884
1952]
1953[Make the case-to-let transformation a little less eager
1954simonpj@microsoft.com**20101208172251
1955 Ignore-this: 55eaa1b5753af31aeb32ec792cb6b662
1956 
1957 See Note [Case elimination: lifted case].
1958 Thanks to Roman for identifying this case.
1959]
1960[warning fix: don't redefine BLOCKS_PER_MBLOCK
1961Simon Marlow <marlowsd@gmail.com>**20101210094002
1962 Ignore-this: cadba57f1c38f5e2af1de37d0a79c7ee
1963]
1964[Only reset the event log if logging is turned on (addendum to #4512)
1965Simon Marlow <marlowsd@gmail.com>**20101210093951
1966 Ignore-this: c9f85f0de2b11a37337672fba59aecc6
1967]
1968[allocate enough room for the longer filename (addendum to #4512)
1969Simon Marlow <marlowsd@gmail.com>**20101210093906
1970 Ignore-this: 270dc0219d98f1e0f9e006102ade7087
1971]
1972[Fix Windows build: move rtsTimerSignal to the POSIX-only section
1973Simon Marlow <marlowsd@gmail.com>**20101210090045
1974 Ignore-this: aa1844b70b9f1a44447787c4bbe10d44
1975]
1976[Default the value of -dppr-cols when the static flags aren't initialised yet
1977Ben Lippmeier <benl@ouroborus.net>**20101210060154
1978 Ignore-this: 4cea29085ef904f379a8829714c9e180
1979 If GHC's command line options are bad then the options parser uses the
1980 pretty printer before the -dppr-cols flag has been read.
1981]
1982[Defensify naked read in LLVM mangler
1983Ben Lippmeier <benl@ouroborus.net>**20101210045922
1984 Ignore-this: 1373f597863851bd03e7a7254558ed04
1985]
1986[Formatting only
1987Ben Lippmeier <benl@ouroborus.net>**20101210042600
1988 Ignore-this: 20bbcd95c70b59094d0bb8a63e459103
1989]
1990[Always ppr case alts on separate lines
1991Ben Lippmeier <benl@ouroborus.net>**20101208070508
1992 Ignore-this: 7e2edd57a61427621aeb254aef84f0f7
1993]
1994[Add -dppr-colsN to set width of dumps
1995Ben Lippmeier <benl@ouroborus.net>**20101208070245
1996 Ignore-this: edc64fee6c373b895bb80b83b549ce1a
1997]
1998[Add -dppr-case-as-let to print "strict lets" as actual lets
1999Ben Lippmeier <benl@ouroborus.net>**20101208065548
2000 Ignore-this: eb1d122dbd73b5263cae3a9f8259a838
2001]
2002[Suppress more info with -dsuppress-idinfo
2003Ben Lippmeier <benl@ouroborus.net>**20101208063037
2004 Ignore-this: 5e8213d7b8d2905e245917aa3e83efc5
2005]
2006[Implement -dsuppress-type-signatures
2007Ben Lippmeier <benl@ouroborus.net>**20101208062814
2008 Ignore-this: 34dbefe5f8d7fe58ecb26d6a748d1c71
2009]
2010[Add more suppression flags
2011Ben Lippmeier <benl@ouroborus.net>**20101208020723
2012 Ignore-this: b010ba9789a2fde6b815f33494fcc23c
2013  -dsuppress-all
2014  -dsuppress-type-applications
2015  -dsuppress-idinfo
2016]
2017[fix ticket number (#4505)
2018Simon Marlow <marlowsd@gmail.com>**20101209120404
2019 Ignore-this: 5769c5ce2a8d69d62d977a9ae138ec23
2020]
2021[fix warnings
2022Simon Marlow <marlowsd@gmail.com>**20101209115844
2023 Ignore-this: ffff37feb2abbfc5bd12940c7007c208
2024]
2025[Catch too-large allocations and emit an error message (#4505)
2026Simon Marlow <marlowsd@gmail.com>**20101209114005
2027 Ignore-this: c9013ab63dd0bd62ea045060528550c6
2028 
2029 This is a temporary measure until we fix the bug properly (which is
2030 somewhat tricky, and we think might be easier in the new code
2031 generator).
2032 
2033 For now we get:
2034 
2035 ghc-stage2: sorry! (unimplemented feature or known bug)
2036   (GHC version 7.1 for i386-unknown-linux):
2037         Trying to allocate more than 1040384 bytes.
2038 
2039 See: http://hackage.haskell.org/trac/ghc/ticket/4550
2040 Suggestion: read data from a file instead of having large static data
2041 structures in the code.
2042]
2043[Export the value of the signal used by scheduler (#4504)
2044Dmitry Astapov <dastapov@gmail.com>**20101208183755
2045 Ignore-this: 427bf8c2469283fc7a6f759440d07d87
2046]
2047[Tweak the "sorry" message a bit
2048Simon Marlow <marlowsd@gmail.com>**20101208163212
2049 Ignore-this: aa1ce5bc3c27111548204b740572efbe
2050 
2051 -              "sorry! (this is work in progress)\n"
2052 +              "sorry! (unimplemented feature or known bug)\n"
2053]
2054[:unset settings support
2055Boris Lykah <lykahb@gmail.com>**20101123190132
2056 Ignore-this: 5e97c99238f5d2394592858c34c004d
2057 Added support for settings [args, prog, prompt, editor and stop].
2058 Now :unset supports the same set of options as :set.
2059]
2060[Fix Windows memory freeing: add a check for fb == NULL; fixes trac #4506
2061Ian Lynagh <igloo@earth.li>**20101208152349
2062 Also added a few comments, and a load of code got indented 1 level deeper.
2063]
2064[Fixes for #4512: EventLog.c - provides ability to terminate event logging, Schedule.c - uses them in forkProcess.
2065Dmitry Astapov <dastapov@gmail.com>**20101203133950
2066 Ignore-this: 2da7f215d6c22708a18291a416ba8881
2067]
2068[Make CPPFLAGS variables, as well as CFLAGS and LDFLAGS
2069Ian Lynagh <igloo@earth.li>**20101207010033
2070 Ignore-this: 2fc1ca1422aae1988d0fe1d29a8485d9
2071 This fixes the "does unsetenv return void" test in the unix package on
2072 OS X, if I tell it to make 10.4-compatible binaries. The test uses
2073 CPPFLAGS but not CFLAGS, so it thought it returned int (as it was
2074 in 10.5-mode), but the C compiler (using CFLAGS, so in 10.4 mode)
2075 thought it returned void.
2076 
2077 I also added CONF_LD_OPTS_STAGE$3 to the list of things in LDFLAGS,
2078 which looks like an accidental ommission.
2079]
2080[Add a configure message
2081Ian Lynagh <igloo@earth.li>**20101206215201]
2082[Link even programs containing no Haskell modules with GHC
2083Ian Lynagh <igloo@earth.li>**20101206203329
2084 I don't remember why we made it use gcc instead, but going back to
2085 using ghc doesn't seem to break anything, and should fix the build
2086 on OS X 10.6.
2087]
2088[Correct the stage that the includes/ tools are built in
2089Ian Lynagh <igloo@earth.li>**20101206203125]
2090[Tweak the cleaning of inplace/; fixes trac #4320
2091Ian Lynagh <igloo@earth.li>**20101205212048]
2092[Close .ghci files after reading them; fixes trac #4487
2093Ian Lynagh <igloo@earth.li>**20101205205301]
2094[Fix the behaviour of :history for ticks surrounding top level functions
2095pepeiborra@gmail.com**20101203202346
2096 Ignore-this: 8059d4859c52c0c9a235b937cb8cde1d
2097]
2098[Don't warn of duplicate exports in case of module exports.
2099Michal Terepeta <michal.terepeta@gmail.com>**20101127212116
2100 Ignore-this: ea225d517826f971c400bbb68d1405b8
2101 
2102 But only when the module exports refer to different modules.
2103 See ticket #4478.
2104]
2105[Fix whitespace/layout in RnNames.
2106Michal Terepeta <michal.terepeta@gmail.com>**20101030171303
2107 Ignore-this: 707a7955fc4fc51683cc5a1dfe57f93
2108]
2109[Tell gcc to support back to OS X 10.5
2110Ian Lynagh <igloo@earth.li>**20101203201558
2111 Ignore-this: f02d70e5b9cce50137981c6cb2b62a18
2112]
2113[Make RelaxedLayout off by default
2114Ian Lynagh <igloo@earth.li>**20101202140808
2115 I suspect this is a vary rarely used extension to the official layout
2116 rule.
2117]
2118[Fix up TcInstDcls
2119simonpj@microsoft.com**20101203180758
2120 Ignore-this: 9311aeb4ee67c799704afec90b5982d0
2121 
2122 I really don't know how this module got left out of my last
2123 patch, namely
2124   Thu Dec  2 12:35:47 GMT 2010  simonpj@microsoft.com
2125   * Re-jig simplifySuperClass (again)
2126 
2127 I suggest you don't pull either the patch above, or this
2128 one, unless you really have to.  I'm not fully confident
2129 that it works properly yet.  Ran out of time. Sigh.
2130]
2131[throwTo: report the why_blocked value in the barf()
2132Simon Marlow <marlowsd@gmail.com>**20101203094840
2133 Ignore-this: 3b167c581be1c51dfe3586cc6359e1d0
2134]
2135[handle ThreadMigrating in throwTo() (#4811)
2136Simon Marlow <marlowsd@gmail.com>**20101203094818
2137 Ignore-this: 8ef8cb7fd3b50a27f83c29968131d461
2138 If a throwTo targets a thread that has just been created with
2139 forkOnIO, then it is possible the exception strikes while the thread
2140 is still in the process of migrating.  throwTo() didn't handle this
2141 case, but it's fairly straightforward.
2142]
2143[removeThreadFromQueue: stub out the link field before returning (#4813)
2144Simon Marlow <marlowsd@gmail.com>**20101202160838
2145 Ignore-this: 653ae17bc1120d7f4130da94665002a1
2146]
2147[small tidyup
2148Simon Marlow <marlowsd@gmail.com>**20101126140620
2149 Ignore-this: 70b1d5ed4c81a7b29dd5980a2d84aae1
2150]
2151[Fix a recomp bug: make classes/datatypes depend directly on DFuns (#4469)
2152Simon Marlow <marlowsd@gmail.com>**20101202122349
2153 Ignore-this: 61c765583bb1d97caa88cf9b4f45b87c
2154 And remove the old mechanism of recording dfun uses separately,
2155 because it didn't work.
2156 
2157 This wiki page describes recompilation avoidance and fingerprinting.
2158 I'll update it to describe the new method and what went wrong with the
2159 old method:
2160 
2161 http://hackage.haskell.org/trac/ghc/wiki/Commentary/Compiler/RecompilationAvoidance
2162]
2163[make a panic message more informative and suggest -dcore-lint (see #4534)
2164Simon Marlow <marlowsd@gmail.com>**20101201151706
2165 Ignore-this: 2a10761925d6f9f52675948baa30f7a
2166]
2167[Re-jig simplifySuperClass (again)
2168simonpj@microsoft.com**20101202123547
2169 Ignore-this: fe4062b8988258f6748ebd8fbd6515b5
2170 
2171 This fixes the current loop in T3731, and will fix other
2172 reported loops.  The loops show up when we are generating
2173 evidence for superclasses in an instance declaration.
2174 
2175 The trick is to make the "self" dictionary simplifySuperClass
2176 depend *explicitly* on the superclass we are currently trying
2177 to build.  See Note [Dependencies in self dictionaries] in TcSimplify.
2178 
2179 That in turn means that EvDFunApp needs a dependency-list, used
2180 when chasing dependencies in isGoodRecEv.
2181]
2182[A little refactoring (remove redundant argument passed to isGoodRecEv)
2183simonpj@microsoft.com**20101202123110
2184 Ignore-this: e517c5c12109a230f08dafb4d1e386df
2185]
2186[Make rebindable if-then-else a little more permissive
2187simonpj@microsoft.com**20101202122540
2188 Ignore-this: ddb552cfe307607b42d1e4baf4e3bf21
2189 
2190 See Note [Rebindable syntax for if].  Fixes Trac #4798.
2191 Thanks to Nils Schweinsberg <mail@n-sch.de>
2192]
2193[Improve error message (Trac #4799)
2194simonpj@microsoft.com**20101202102706
2195 Ignore-this: d9896e4d182936de1f256c820b96a8cf
2196]
2197[Fix a nasty bug in RULE matching: Trac #4814
2198simonpj@microsoft.com**20101202102618
2199 Ignore-this: ba058ad46a02bd2faf3a14de93fd19c6
2200 
2201 See Note [Matching lets], which explains it all in detail.
2202 It took me a day to think of a nice way to fix the bug,
2203 but I think the result is quite respectable. Subtle, though.
2204]
2205[Rename -XPArr to -XParallelArrays
2206Ben Lippmeier <benl@ouroborus.net>**20101130075415
2207 Ignore-this: 21b37680a7f25800d1200b59ad0b6b39
2208]
2209[FIX #1845 (unconditional relative branch out of range)
2210pho@cielonegro.org**20101130143014
2211 Ignore-this: df234bd8ad937104c455656fe3c33732
2212 
2213 Don't use mmap on powerpc-apple-darwin as mmap doesn't support
2214 reallocating but we need to allocate jump islands just after each
2215 object images. Otherwise relative branches to jump islands can fail
2216 due to 24-bits displacement overflow.
2217]
2218[rts/Linker.c (loadArchive):
2219pho@cielonegro.org**20101130142700
2220 Ignore-this: bc84f9369ce5c2d289440701b7a3a2ab
2221 
2222 This routine should be aware of Mach-O misalignment of malloc'ed memory regions.
2223]
2224[rts/Linker.c (machoGetMisalignment):
2225pho@cielonegro.org**20101130123355
2226 Ignore-this: 75425600049efd587e9873578e26392f
2227 
2228 Use fseek(3) instead of rewind(3) to move the file position indicator back to the initial position. Otherwise we can't use this function in loadArchive().
2229]
2230[rts/Linker.c (ocFlushInstructionCache):
2231pho@cielonegro.org**20101130121425
2232 Ignore-this: 1e2c207e4b1d17387617ec5d645204b7
2233 
2234 I found this function causes a segfault when ocAllocateSymbolExtras() has allocated a separate memory region for jump islands.
2235]
2236[Remove NewQualifiedOperators
2237Ian Lynagh <igloo@earth.li>**20101201181117
2238 The extension was rejected by Haskell', and deprecated in 7.0.
2239]
2240[fix ref to utils/ext-core, which moved to Hackage (extcore package)
2241Simon Marlow <marlowsd@gmail.com>**20101201092147
2242 Ignore-this: 272a7daaa335ef60bcc645db70b4d68b
2243]
2244[fix floating-point/FFI section: fenv is C99, not POSIX
2245Simon Marlow <marlowsd@gmail.com>**20101201092119
2246 Ignore-this: ce8b3edd428e4f77691dd739b5b4ae73
2247]
2248[Fixed some 'unused vars' warnings
2249keller@cse.unsw.edu.au**20101130013425
2250 Ignore-this: 35790d443faa23b87e4ba442e62376a3
2251]
2252[vectScalarLam handles int, float, and double now
2253keller@cse.unsw.edu.au**20101129231043
2254 Ignore-this: 6d67bdc8dd8577184040e791e6f3d0
2255]
2256[Handling of lets, letrec and case when checking if a lambda expr needs to be vectorised
2257keller@cse.unsw.edu.au**20101115051225
2258 Ignore-this: 1db6ed63d7b3f6d093e019322b407ff7
2259]
2260[Document the behaviour of fenv.h functions with GHC (#4391)
2261Simon Marlow <marlowsd@gmail.com>**20101126125336
2262 Ignore-this: bc4eab49428d567505a28add6fed90f1
2263]
2264[Remove the no-ghci-lib warning in ghc-pkg
2265Ian Lynagh <igloo@earth.li>**20101127235805
2266 GHCi libs are no longer necessary, as we can use the .a or .so versions
2267 instead.
2268]
2269[Add GNU-variant support to the .a parser, and other improvements/tidyups
2270Ian Lynagh <igloo@earth.li>**20101127223945]
2271[Re-indent only
2272Ian Lynagh <igloo@earth.li>**20101127191646]
2273[Improve linker debugging for archive files
2274Ian Lynagh <igloo@earth.li>**20101127190907]
2275[Always enable the archive-loading code
2276Ian Lynagh <igloo@earth.li>**20101127173000
2277 If the GHCi .o lib doesn't exist, load the .a instead
2278]
2279[Inherit the ForceSpecConstr flag in non-recursive nested bindings
2280Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101127125025
2281 Ignore-this: 401391eae25cefcb4afaba2e357decc1
2282 
2283 This makes sure that join points are fully specialised in loops which are
2284 marked as ForceSpecConstr.
2285]
2286[Document -ddump-rule-firings and -ddump-rule-rewrites
2287Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101127123528
2288 Ignore-this: beade2efe0cd767c0ce9d4f45a3380ba
2289]
2290[New flag -dddump-rule-rewrites
2291Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101127122022
2292 Ignore-this: c0ef5b8a199fbd1ef020258d2cde85a3
2293 
2294 Now, -ddump-rule-firings only shows the names of the rules that fired (it would
2295 show "before" and "after" with -dverbose-core2core previously) and
2296 -ddump-rule-rewrites always shows the "before" and "after" bits, even without
2297 -dverbose-core2core.
2298]
2299[Acutally, wild-card variables *can* have occurrences
2300simonpj@microsoft.com**20101126162409
2301 Ignore-this: 544bffed75eeccef03a1097f98524eea
2302 
2303 This patch removes the Lint test, and comments why
2304]
2305[Tidy up the handling of wild-card binders, and make Lint check it
2306simonpj@microsoft.com**20101126133210
2307 Ignore-this: 9e0be9f7867d53046ee5b0e478a0f433
2308 
2309 See Note [WildCard binders] in SimplEnv.  Spotted by Roman.
2310]
2311[Substitution should just substitute, not optimise
2312simonpj@microsoft.com**20101125172356
2313 Ignore-this: 657628d9b6796ceb5f915c43d56e4a06
2314 
2315 This was causing Trac #4524, by optimising
2316      (e |> co)  to   e
2317 on the LHS of a rule.  Result, the template variable
2318 'co' wasn't bound any more.
2319 
2320 Now that substition doesn't optimise, it seems sensible to call
2321 simpleOptExpr rather than substExpr when substituting in the
2322 RHS of rules.  Not a big deal either way.
2323]
2324[Make SpecConstr "look through" identity coercions
2325simonpj@microsoft.com**20101125172138
2326 Ignore-this: c1cc585ed890a7702c33987e971e0af6
2327]
2328[Comment only
2329simonpj@microsoft.com**20101125172011
2330 Ignore-this: 3c7be8791badd00dcca9610ebb8981d1
2331]
2332[White space only
2333simonpj@microsoft.com**20101101080748
2334 Ignore-this: f7133fc6b22ae263c6672543a8534a6f
2335]
2336[Keep a maximum of 6 spare worker threads per Capability (#4262)
2337Simon Marlow <marlowsd@gmail.com>**20101125135729
2338 Ignore-this: a020786569656bf2f3a1717b65d463bd
2339]
2340[Unicide OtherNumber category should be allowed in identifiers (#4373)
2341Simon Marlow <marlowsd@gmail.com>**20101115095444
2342 Ignore-this: e331b6ddb17550163ee91bd283348800
2343]
2344[vectoriser: fix warning
2345Ben Lippmeier <benl@ouroborus.net>**20101126044036
2346 Ignore-this: e1a66bb405bf2f3f56b42c3b13fd4bf3
2347]
2348[vectoriser: fix warning
2349Ben Lippmeier <benl@ouroborus.net>**20101126042950
2350 Ignore-this: df8dd25bcfb3946c2974b13953a2f2c7
2351]
2352[vectoriser: take class directly from the instance tycon
2353Ben Lippmeier <benl@ouroborus.net>**20101126042900
2354 Ignore-this: 626a416717a5a059f39e53f4ec95fc66
2355]
2356[vectoriser: comments only
2357Ben Lippmeier <benl@ouroborus.net>**20101125073201
2358 Ignore-this: 8846ea8895307083bd1ebbc5d7fb1c5
2359]
2360[vectoriser: follow changes in mkClass
2361Ben Lippmeier <benl@ouroborus.net>**20101125062349
2362 Ignore-this: d5018cc022686d4272e126ca9a12283a
2363]
2364[vectoriser: tracing wibbles
2365Ben Lippmeier <benl@ouroborus.net>**20101125062332
2366 Ignore-this: c2024d8f03bc03bee2851f4f1c139fd5
2367]
2368[mkDFunUnfolding wants the type of the dfun to be a PredTy
2369benl@ouroborus.net**20100914062939
2370 Ignore-this: 7aa6e6b140746184cf00355b50c83b66
2371]
2372[vectoriser: fix conflicts
2373Ben Lippmeier <benl@ouroborus.net>**20101125060904
2374 Ignore-this: cc3decab1affada8629ca3818b76b3bf
2375]
2376[Comments and formatting only
2377benl@ouroborus.net**20100914062903
2378 Ignore-this: b0fc25f0952cafd56cc25353936327d4
2379]
2380[Comments and formatting to type environment vectoriser
2381benl@ouroborus.net**20100909080405
2382 Ignore-this: ab8549d53f845c9d82ed9a525fda3906
2383]
2384[Don't mix implicit and explicit layout
2385Ian Lynagh <igloo@earth.li>**20101124231514]
2386[Whitespace only
2387Ian Lynagh <igloo@earth.li>**20101124230655]
2388[Separate NondecreasingIndentation out into its own extension
2389Ian Lynagh <igloo@earth.li>**20101124220507]
2390[Add another GHC layout rule relaxation to RelaxedLayout
2391Ian Lynagh <igloo@earth.li>**20101124205957]
2392[Remove an unused build system variable: GhcDir
2393Ian Lynagh <igloo@earth.li>**20101124140455]
2394[Remove unused build system variable: GhcHasEditline
2395Ian Lynagh <igloo@earth.li>**20101124140415]
2396[Remove unused variables from the build system: HBC, NHC, MKDEPENDHS
2397Ian Lynagh <igloo@earth.li>**20101124140052]
2398[Remove references to Haskell 98
2399Ian Lynagh <igloo@earth.li>**20101123233536
2400 They are no longer right, as we have Haskell' generating new Haskell
2401 standards.
2402]
2403[Tweak a configure test
2404Ian Lynagh <igloo@earth.li>**20101123170621]
2405[Add a configure test for the visibility hidden attribute
2406Ian Lynagh <igloo@earth.li>**20101123170541]
2407[sanity: fix places where we weren't filling fresh memory with 0xaa
2408Simon Marlow <marlowsd@gmail.com>**20101029092843
2409 Ignore-this: 2cb18f7f5afcaf33371aeffce67e218f
2410]
2411[Just some alpha renaming
2412Ian Lynagh <igloo@earth.li>**20101121144455
2413 Ignore-this: d5e807c5470840efc199e29f7d50804c
2414]
2415[Fix bug #3165 (:history throws irrefutable pattern failed)
2416pepeiborra@gmail.com**20101115223623
2417 Ignore-this: 73edf56e502b4d0385bc044133b27946
2418 
2419 I ran across this bug and took the time to fix it, closing
2420 a long time due TODO in InteractiveEval.hs
2421 
2422 Instead of looking around to find the enclosing declaration
2423 of a tick, this patch makes use of the information already collected during the
2424 coverage desugaring phase
2425]
2426[For bindists, build ghc-pwd with stage 1
2427Ian Lynagh <igloo@earth.li>**20101121183520
2428 Ignore-this: a3b5c8b78c81ec1b6d5fbf23da346ff5
2429 rather then the bootstrapping compiler. This fixes problems where the
2430 bootstrapping compiler dynamically links against libraries not on the
2431 target machine.
2432]
2433[Makefile tweak
2434Ian Lynagh <igloo@earth.li>**20101121183342
2435 Ignore-this: cd55a2819c1a5fd36da1bc7a75d2ded1
2436]
2437[Fix a makefile include ordering sanity check
2438Ian Lynagh <igloo@earth.li>**20101121174916
2439 Ignore-this: d0bdd41c4b618944d04ecb4f54fdd0f1
2440]
2441[Add an extension for GHC's layout-rule relaxations
2442Ian Lynagh <igloo@earth.li>**20101120215340
2443 Still TODO: Add the other relaxation (#1060) and update the alternative
2444 layout rule to use the extension.
2445]
2446[Tweak the bindist configure.ac.in
2447Ian Lynagh <igloo@earth.li>**20101120173735]
2448[configure.ac tweaks
2449Ian Lynagh <igloo@earth.li>**20101120170245]
2450[When testing the bindist, tell it where gcc is
2451Ian Lynagh <igloo@earth.li>**20101120155920
2452 The location isn't baked into the bindist, as it may differ from
2453 machine to machine.
2454]
2455[Comments only
2456simonpj@microsoft.com**20101119100153
2457 Ignore-this: 7abd5d965ea805770449d6f8dadbb921
2458]
2459[ForceSpecConstr now forces specialisation even for arguments which aren't scrutinised
2460Roman Leshchinskiy <rl@cse.unsw.edu.au>**20101118212839
2461 Ignore-this: db45721d29a694e53746f8b76513efa4
2462]
2463[Move the superclass generation to the canonicaliser
2464simonpj@microsoft.com**20101118120533
2465 Ignore-this: 5e0e525402a240b709f2b8104c1682b2
2466 
2467 Doing superclass generation in the canonicaliser (rather than
2468 TcInteract) uses less code, and is generally more efficient.
2469 
2470 See Note [Adding superclasses] in TcCanonical.
2471 
2472 Fixes Trac #4497.
2473]
2474[Fix the generation of in-scope variables for IfaceLint check
2475simonpj@microsoft.com**20101118090057
2476 Ignore-this: bbcdba61ddf89d07fe69ca99c2017e3f
2477]
2478[Comments only
2479simonpj@microsoft.com**20101118090034
2480 Ignore-this: fa2936d35a0f7be4e4535ea9e2b7bf7b
2481]
2482[Omit bogus test for -XDeriveFunctor
2483simonpj@microsoft.com**20101118090028
2484 Ignore-this: a534243011809ebbb788b910961601c5
2485 
2486 It was duplicated in the case of 'deriving( Functor )'
2487 and wrong for 'deriving( Foldable )'
2488]
2489[Improve error message on advice from a user
2490simonpj@microsoft.com**20101118085306
2491 Ignore-this: bd4f3858ff24e602e985288f27d536f3
2492 
2493 See Trac #4499
2494]
2495[TAG 2010-11-18
2496Ian Lynagh <igloo@earth.li>**20101118011554
2497 Ignore-this: ccadbe7fadd1148d2ee3caa8c8821ec5
2498]
2499Patch bundle hash:
2500d25d090c9303e567f2a326aa9d601391ab839ee1