Ticket #2603: korn_shell.patch

File korn_shell.patch, 9.9 KB (added by NeilMitchell, 6 years ago)
Line 
1
2New patches:
3
4[Make the matching of the filename ghc.exe case insensitive, fixes bug #2603
5Neil Mitchell <[email protected]>**20080916160311] {
6hunk ./compiler/main/SysTools.lhs 846
7-                (d, "ghc.exe") ->
8+                (d, ghc_exe) | map toLower ghc_exe == "ghc.exe" ->
9}
10
11Context:
12
13[Stop using mremap() to allocate space for trampolines
14Simon Marlow <[email protected]>**20080915145924
15 
16 This was causing problems because sometimes mremap() moved the memory
17 we had allocated from the low 2Gb to above the 2Gb boundary, causing
18 some linkages to fail.  There's no MAP_32BIT flag to mremap().
19 
20 So now we just use mmap(MAP_ANON|MAP_32BIT) to allocated space for the
21 trampolines.  People without MAP_32BIT (eg. *BSD) will still have to
22 do something else here, such as allocating memory from a fixed
23 address; so I've made it slightly easier for those guys, but there's
24 still work to do (#2063).
25 
26 One solution (that Simon PJ is advocating) is to turn on -fPIC by
27 default on x86-64.  This is a good solution as it removes the need for
28 MAP_32BIT, but doesn't work with -fvia-C, so probably this is for
29 later.
30]
31[add $(GhcStage[123]HcOpts)
32Simon Marlow <[email protected]>**20080912155549]
33[Improve handling of -fdph-* flags
34Roman Leshchinskiy <[email protected]>**20080916034746]
35[Add -fdph-this
36Roman Leshchinskiy <[email protected]>**20080916033710]
37[Remove last traces of package ndp
38Roman Leshchinskiy <[email protected]>**20080916033428]
39[Clean up vectorisation error messages
40Roman Leshchinskiy <[email protected]>**20080916013236]
41[Fix vectoriser bug
42Roman Leshchinskiy <[email protected]>**20080915042823
43 
44 We were using mkWildId in situations where it cause disastrous shadowing
45]
46[Track changes to dph
47Roman Leshchinskiy <[email protected]>**20080912114028]
48[Change desugaring of PArr literals
49Roman Leshchinskiy <[email protected]>**20080912015609]
50[Expose the dph packages automatically if -dph-* is set
51Roman Leshchinskiy <[email protected]>**20080912004403]
52[Don't panic on non-vectorisable expressions
53Roman Leshchinskiy <[email protected]>**20080911054333]
54[-Odph implies -fno-spec-constr-count
55Roman Leshchinskiy <[email protected]>**20080910045339]
56[Improve warning for SpecConstr
57[email protected]**20080915154908]
58[White space only
59[email protected]**20080915154841]
60[Minor refactoring to get rid of Type.splitNewTyConApp
61[email protected]**20080915072946]
62[Refactor the desugaring of RULE lhss a bit
63[email protected]**20080915150601
64 
65 This is just a tidy-up.  Previously we were calling occurAnalyse
66 twice on each LHS which was silly and a bit unclean too. 
67 
68 This patch should have no overall effect, though.
69 
70]
71[Do not use the Static Arg Transformation by default with -O2
72[email protected]**20080915150433
73 
74 Max has some more heuristics to add, and is meanwhile worried
75 that having SAT on by default will make some highly-tuned array
76 programs worse.  So it's off by default. 
77 
78 Use -fstatic-argument-transformation to switch it on.
79 
80]
81[Comments, and a couple of asserts, only
82[email protected]**20080914114641]
83[Fix Trac #2587: take account of type lets
84[email protected]**20080914113434
85 
86 GHC allows a non-recursive let for type varaibles
87        let a = TYPE ty in ...
88 But the free-variable finder had not caught up with this
89 fact. This patch catches up.
90 
91]
92[Don't try to float type-lets
93[email protected]**20080914113324
94 
95 A type let shouldn't really occur in SetLevels, but if it does,
96 this patch makes sure it is left alone.
97 
98]
99[add refs and fix a bug (noted by Peter Gammie) in docs of arrow notation
100Ross Paterson <[email protected]>**20080915104757]
101[Generalise type of 'withExtendedLinkEnv'.
102Thomas Schilling <[email protected]>**20080915085738]
103[Use 'GhcMonad' in ghci/InteractiveUI.
104Thomas Schilling <[email protected]>**20080915085633]
105[Use 'GhcMonad' in ghci/GhciTags.
106Thomas Schilling <[email protected]>**20080915084922]
107[Use 'GhcMonad' in ghci/Debugger.
108Thomas Schilling <[email protected]>**20080915084738]
109[Use 'GhcMonad' in ghci/GhciMonad.
110Thomas Schilling <[email protected]>**20080915084646]
111[Use 'GhcMonad' in ghc/Main.
112Thomas Schilling <[email protected]>**20080914232957]
113[Require PatternSignatures for bootstrapping compiler.
114Thomas Schilling <[email protected]>**20080914232642]
115[Use 'GhcMonad' in InteractiveEval.
116Thomas Schilling <[email protected]>**20080914232454]
117[Use 'GhcMonad' in GHC and split up checkModule into phases.
118Thomas Schilling <[email protected]>**20080914232044
119 
120 I'm not sure I covered all the ways of throwing errors in the code.
121 Some functions throw ProgramErrors, some error messages.  It's still
122 quite a mess, but we're getting closer.  The missing cases are mostly
123 errors that cannot be fixed by the API client either or are a result
124 of wrong usage, so are in any case fatal.
125 
126 One function, 'getModuleInfo', still returns a 'Maybe', but the
127 documentation suggests it should always succeed.  So I may change that
128 soon.
129 
130 The spit-up of of 'checkModule' has pros and cons.  The various forms
131 of 'checkModule*' now become:
132 
133  checkAndLoadModule ms False ~~>
134     loadModule =<< typecheckModule =<< parseModule (ms_mod_name ms)
135 
136  checkAndLoadModule ms True ~~>
137    loadModule =<< desugarModule =<< typecheckModule =<< parseModule (ms_mod_name ms)
138 
139  checkModule mn False ~~>
140    typecheckModule =<< parseModule mn
141 
142  checkModule mn True ~~>
143    desugarModule =<< typecheckModule =<< parseModule mn
144 
145 The old APIs cannot easily be provided, since the result type would be
146 different depending on the second argument.  However, a more
147 convenient API can be modelled on top of these four functions
148 ({parse,typecheck,desugar,load}Module).
149]
150[Use 'GhcMonad' in DriverPipeline.  Also haddockify a bit while we're at it.
151Thomas Schilling <[email protected]>**20080914220628]
152[Use 'GhcMonad' in HscMain.
153Thomas Schilling <[email protected]>**20080914213655]
154[Use 'GhcMonad' in DriverMkDepend.
155Thomas Schilling <[email protected]>**20080914212113]
156[Haddockify DynFlags (partial).
157Thomas Schilling <[email protected]>**20080914211718]
158[Haddockify 'IE'.
159Thomas Schilling <[email protected]>**20080914210016]
160[Provide accessors for 'ImportDecl'.
161Thomas Schilling <[email protected]>**20080914205811]
162[Start haddockifying 'HsBindLR'.
163Thomas Schilling <[email protected]>**20080914205629]
164[Document 'parseStaticFlags'.
165Thomas Schilling <[email protected]>**20080914205316]
166[Introduce 'GhcMonad' class and two default implementations 'Ghc' and 'GhcT'.
167Thomas Schilling <[email protected]>**20080914204930
168 
169 This monad will be required by most public API calls.
170]
171[Give the "Failing due to -Werror" message a name.
172Thomas Schilling <[email protected]>**20080914173904]
173[Make typechecker top-level functions also return messages instead of
174Thomas Schilling <[email protected]>**20080914173228
175 printing them.
176]
177[Reflect changes of desugarer error reporting in VectMonad.
178Thomas Schilling <[email protected]>**20080914172711]
179[Generalise 'handleGhcException' to work with any 'ExceptionMonad'.
180Thomas Schilling <[email protected]>**20080914172404]
181[Introduce an 'ExceptionMonad' class.
182Thomas Schilling <[email protected]>**20080914172154
183 
184 This monad provides variants of 'catch', 'bracket', and 'finally', so
185 exceptions can be handled in monads that wrap IO.  The latter two
186 methods need to be part of the class definition, because GHC uses
187 'block' and 'unblock' which are used in the definition of those two
188 methods for the IO monad.  A perhaps better class interface would
189 consist of 'gcatch', 'gblock', and 'gunblock' and let the latter two
190 default to 'id' like is done for non-GHC implementations of 'bracket'
191 and 'finally'.
192]
193[Provide default MonadIO instance for IO.
194Thomas Schilling <[email protected]>**20080914164245]
195[Return instead of print warnings and errors in desugarer.
196Thomas Schilling <[email protected]>**20080914163641]
197[Return parser errors and warnings instead of dying.
198Thomas Schilling <[email protected]>**20080914162644]
199[Add aliases for bags of warnings and errors.
200Thomas Schilling <[email protected]>**20080914160337]
201[Slightly more helpful panic message in DynFlags
202Tim Chevalier <[email protected]>**20080915080650]
203[Comments only: ".core" => ".hcr"
204Tim Chevalier <[email protected]>**20080914203645]
205[We need to tell ghc-pkg to --force if we've only built a profiling library
206Ian Lynagh <[email protected]>**20080913153142]
207[If we're profiling GHC, don't bother building the GHC package the vanilla way
208Ian Lynagh <[email protected]>**20080913144820]
209[Remove the duplicate show rule in libraries/Makefile
210Ian Lynagh <[email protected]>**20080913144413]
211[Move the "show" target from target.mk to boilerplate.mk
212Ian Lynagh <[email protected]>**20080913141312
213 target.mk isn't included everywhere, but show is always handy
214]
215[Change how we detect if we are using the bootstrapping compiler or not
216Ian Lynagh <[email protected]>**20080913104658
217 I think looking for $(GHC_COMPILER_DIR_ABS) was failing on the Windows
218 buildbot due to different path separators. Now we just look for
219 "inplace".
220]
221[wibble the distrib Makefile
222Ian Lynagh <[email protected]>**20080912135932
223 We now need to install driver after ghc
224]
225[Reinstate the driver/ghc directory, to create a versioned GHC program
226Ian Lynagh <[email protected]>**20080912113619
227 e.g. $(bindir)/ghc-6.9.20080911
228]
229[If USE_NEW_MKDEPEND_FLAGS is YES then don't use the deprecated -optdep flags
230Ian Lynagh <[email protected]>**20080912110316]
231[Use --force-local when calling tar in bindisttest/
232Ian Lynagh <[email protected]>**20080912012855
233 Otherwise it thinks that c:/foo is a remote file
234]
235[Fix #2586, bug in THUNK_SELECTORs (again)
236Simon Marlow <[email protected]>**20080912130404
237 This time, we had forgetten the write barrier in one place.
238]
239[TAG 2008-09-12 2
240Ian Lynagh <[email protected]>**20080912132848]
241Patch bundle hash:
2421ae5806855a27969217c7431a7a3a8722e699fff