simonmar [Wed, 22 May 2002 09:02:41 +0000 (09:02 +0000)]
[project @ 2002-05-22 09:02:40 by simonmar]
We don't pass --old-ghc to Happy any more. Also, wire $(HappyOpts) to
-agc in config.mk rather than passing it through from configure, since
it was always the same anyhow.
wolfgang [Tue, 21 May 2002 14:58:51 +0000 (14:58 +0000)]
[project @ 2002-05-21 14:58:49 by wolfgang]
Bug-fixes for PowerPC
*) Rewrote StgRun in assembler (that seems to be the cleanest way to get it to work).
*) Fixed a small typo in createAdjustor.
simonpj [Tue, 21 May 2002 13:43:59 +0000 (13:43 +0000)]
[project @ 2002-05-21 13:43:59 by simonpj]
Parse External Core correctly
Amazingly, recursive bindings with only one binding in the
group were being parsed as non-recursive.
ken [Sat, 18 May 2002 05:28:15 +0000 (05:28 +0000)]
[project @ 2002-05-18 05:28:14 by ken]
Add missing va_end's to va_start's.
simonmar [Fri, 17 May 2002 09:03:13 +0000 (09:03 +0000)]
[project @ 2002-05-17 09:03:13 by simonmar]
If we have a Happy version 1.13 or later, turn on the --old-ghc option
for now.
simonpj [Fri, 17 May 2002 08:21:51 +0000 (08:21 +0000)]
[project @ 2002-05-17 08:21:51 by simonpj]
Wibble
simonpj [Fri, 17 May 2002 08:13:08 +0000 (08:13 +0000)]
[project @ 2002-05-17 08:13:08 by simonpj]
Win32 updates from Simon
sof [Wed, 15 May 2002 20:25:10 +0000 (20:25 +0000)]
[project @ 2002-05-15 20:25:10 by sof]
make prev commit work
sof [Wed, 15 May 2002 19:46:11 +0000 (19:46 +0000)]
[project @ 2002-05-15 19:46:11 by sof]
Assume the presence of a Bits Int32 instance only
sof [Wed, 15 May 2002 18:20:28 +0000 (18:20 +0000)]
[project @ 2002-05-15 18:20:28 by sof]
add Main.hi to CLEAN_FILES to cope with the fact that HaskTags.hs is rooming the Main module
chak [Wed, 15 May 2002 09:00:00 +0000 (09:00 +0000)]
[project @ 2002-05-15 08:59:58 by chak]
* Added the options `-ffi' and `-fffi', which switch on FFI support
(`-fglasgow-exts' implies `-fffi').
NB: Strictly speaking `-fffi' is the correct name; however, `-ffi' is
supported as an alternative spelling.
* `-fglasgow-exts' no longer enables the `with' keyword for implicit
parameters. To use `with' as a keyword, the additional option `-fwith' has
to be supplied. (Note that SimonM recently enabled the use of `let' instead
of `with'.)
NB: This might prompt some makefile tweaks in libraries or regression tests.
* Renamed `Foreign.Marshal.Utils.withObject' to `Foreign.Marshal.Utils.with'
as required by the FFI Addendum. (The old name is still available, but
marked as deprecated.)
* Added `realloc' to `Foreign.Marshal.Alloc' (tracking RC4 of the FFI
Addendum).
Docu
~~~~
* Added `-ffi', `-fffi', and `-fwith' to the flag reference and the section
describing options for Glasgow extensions
* Removed most of the FFI section in the User's Guide in favour of a reference
to the Addendum plus a brief description of additional features supported by
GHC.
NB: The old material is still available at fptools/docs/, I merely removed
the reference in the User's Guide.
matthewc [Tue, 14 May 2002 08:30:34 +0000 (08:30 +0000)]
[project @ 2002-05-14 08:30:34 by matthewc]
* Note that ia64-linux works unregisterised and a registerised port is in
progress.
* Recommend bootstrapping from a machine of the same word size :)
matthewc [Tue, 14 May 2002 08:27:11 +0000 (08:27 +0000)]
[project @ 2002-05-14 08:27:11 by matthewc]
Make adjustor unavailability on a particular platform a runtime barf
instead of an #error (to allow the remainder of ghc to function without
architecture-specific support)
matthewc [Tue, 14 May 2002 08:25:46 +0000 (08:25 +0000)]
[project @ 2002-05-14 08:25:46 by matthewc]
Bootstrapping fixes:
* Correct includes and libraries to use new libraries/ tree
* Check UseGhcForCc for "YES" instead of ""
matthewc [Tue, 14 May 2002 08:23:11 +0000 (08:23 +0000)]
[project @ 2002-05-14 08:23:11 by matthewc]
Need #include <stdlib.h> for malloc (otherwise the C compiler will
assume the return value is an int, with bad results)
matthewc [Tue, 14 May 2002 08:17:38 +0000 (08:17 +0000)]
[project @ 2002-05-14 08:17:38 by matthewc]
Round up GMP allocation requests to a whole number of StgWords
(GMP limbs are not necessarily an StgWord in size)
matthewc [Tue, 14 May 2002 08:15:49 +0000 (08:15 +0000)]
[project @ 2002-05-14 08:15:49 by matthewc]
IA64 only changes:
* Place HEAP_BASE in an appropriate place
* On IA64, use GCC "section" attribute to coax compiler into putting info
tables in .text
matthewc [Tue, 14 May 2002 08:09:07 +0000 (08:09 +0000)]
[project @ 2002-05-14 08:09:07 by matthewc]
* Add ia64-*-linux* as a recognised system type
* Standardise capitalisation of YES and NO in HaveLibGmp
* Error out if user tries to use in-tree GMP on IA64 (results in hangs
and segfaults - it probably needs an upgrade sometime)
sof [Sat, 11 May 2002 13:58:18 +0000 (13:58 +0000)]
[project @ 2002-05-11 13:58:18 by sof]
Have createThread() use a separate mutex to ensure unique
allocation of ThreadIds. A less-than-light solution, but
cleaner than trying to reuse sched_mutex for this purpose.
panne [Sat, 11 May 2002 10:02:22 +0000 (10:02 +0000)]
[project @ 2002-05-11 10:02:21 by panne]
Include hschooks.h via a global option to get ghc_strlen's
prototype. It's too omnipresent for per-file OPTIONS.
sof [Sat, 11 May 2002 00:16:12 +0000 (00:16 +0000)]
[project @ 2002-05-11 00:16:11 by sof]
As a result of calling exitScheduler(), enter into a
'shutting-down' state, so that the subsequent (sequential)
execution of finalisers won't get stuck inside the
Scheduler waiting for more work.
Cleanest way I could think of solving this problem on a
Friday afternoon.
sof [Fri, 10 May 2002 22:02:02 +0000 (22:02 +0000)]
[project @ 2002-05-10 22:02:02 by sof]
stopTaskManager(): no seppuku, please.
panne [Fri, 10 May 2002 20:44:39 +0000 (20:44 +0000)]
[project @ 2002-05-10 20:44:29 by panne]
Re-enable bootstrapping: More Ptr trouble, now that it's (almost) abstract...
simonpj [Fri, 10 May 2002 13:34:18 +0000 (13:34 +0000)]
[project @ 2002-05-10 13:34:18 by simonpj]
Remove {| |} syntax from expressions; never fully implemented
simonpj [Fri, 10 May 2002 12:43:02 +0000 (12:43 +0000)]
[project @ 2002-05-10 12:43:02 by simonpj]
If a type signature mentions a type variable that doesn't appear in
the type, GHC was dying horribly. Example (from happy -s):
type T a = () -> ()
f :: T a
f () = ()
This commit fixes the problem.
sof [Wed, 8 May 2002 07:26:36 +0000 (07:26 +0000)]
[project @ 2002-05-08 07:26:36 by sof]
summarise{File}: filter out GHC.Prim from import lists
sof [Tue, 7 May 2002 22:44:41 +0000 (22:44 +0000)]
[project @ 2002-05-07 22:44:41 by sof]
undo the undo; just me being confused.
sof [Tue, 7 May 2002 18:08:24 +0000 (18:08 +0000)]
[project @ 2002-05-07 18:08:24 by sof]
backing out lazyGet'ing of usages; introduces bad breakage w/ GHCi
simonmar [Tue, 7 May 2002 09:47:27 +0000 (09:47 +0000)]
[project @ 2002-05-07 09:47:27 by simonmar]
Mystery at least partially solved: it appears that in the FreeBSD
catalog we must list the iso8859 catalog *before* the docbook catalog,
otherwise things don't work properly.
sof [Tue, 7 May 2002 04:58:15 +0000 (04:58 +0000)]
[project @ 2002-05-07 04:58:15 by sof]
fprintf_ccs(): use snprintf() only if available
sof [Tue, 7 May 2002 00:46:09 +0000 (00:46 +0000)]
[project @ 2002-05-07 00:46:09 by sof]
downsweep.getSummary: keep track of the module a particular ModuleName
stems from, so that we can emit a better error message should the
module be out of scope - e.g.,
sna$ ghc --make -iutil Foo.hs
ghc.exe: chasing modules from: Foo.hs
ghc.exe: can't find module `Xyzz' (while processing "util/Bar.hs")
sna$
simonmar [Mon, 6 May 2002 08:31:29 +0000 (08:31 +0000)]
[project @ 2002-05-06 08:31:29 by simonmar]
Slurp the usages information lazilly from the interface file. In most
cases we don't need it at all: we only use it when we're loading a
module's interface file to check whether the module needs recompiling.
simonmar [Mon, 6 May 2002 08:30:05 +0000 (08:30 +0000)]
[project @ 2002-05-06 08:30:05 by simonmar]
- Make the mi_usages and mi_deprecs fields of ModIface lazy again
(mi_deprecs was already slurped lazilly from the interface file, but
the strictness annotation on its field meant it was actually slurped
strictly. We're about to slurp mi_usages lazilly too.)
panne [Sun, 5 May 2002 17:03:54 +0000 (17:03 +0000)]
[project @ 2002-05-05 17:03:54 by panne]
Simon Sez: "Don't use -s!"
simonmar [Fri, 3 May 2002 13:09:47 +0000 (13:09 +0000)]
[project @ 2002-05-03 13:09:47 by simonmar]
In the get method for FastString, do the dictionary lookup strictly
(speeds things up and reduces the residency by a few k).
simonmar [Fri, 3 May 2002 11:25:58 +0000 (11:25 +0000)]
[project @ 2002-05-03 11:25:58 by simonmar]
seqList the result of mkImportInfo: this fixes a space leak whereby
the entire collection of Ifaces read by the renamer were being held on
to during the core-to-core phases.
simonmar [Fri, 3 May 2002 11:23:41 +0000 (11:23 +0000)]
[project @ 2002-05-03 11:23:41 by simonmar]
I fluffed the seqList call in the previous revision, so it wasn't
actually doing anything. I've verified that fixing it properly *does*
remove a space leak.
simonmar [Fri, 3 May 2002 11:10:47 +0000 (11:10 +0000)]
[project @ 2002-05-03 11:10:47 by simonmar]
Eliminate a duplicate binding
simonmar [Thu, 2 May 2002 14:37:27 +0000 (14:37 +0000)]
[project @ 2002-05-02 14:37:27 by simonmar]
Overhaul:
- Fix the markup in various ways, and indent it so it doesn't look
so much like it was machine-generated by a badly written perl
script
- Add a section entitled "Porting GHC", describing the intricacies
of porting from unregisterised .hc files, amongst other things.
Comments/corrections welcome.
simonmar [Thu, 2 May 2002 13:33:27 +0000 (13:33 +0000)]
[project @ 2002-05-02 13:33:27 by simonmar]
Add this script which does an unregisterised bootstrap to build HC
files for porting GHC.
simonmar [Thu, 2 May 2002 09:09:18 +0000 (09:09 +0000)]
[project @ 2002-05-02 09:09:18 by simonmar]
ptext => ftext
sof [Wed, 1 May 2002 22:17:31 +0000 (22:17 +0000)]
[project @ 2002-05-01 22:17:29 by sof]
check for <sys/uio.h>
sof [Wed, 1 May 2002 22:16:30 +0000 (22:16 +0000)]
[project @ 2002-05-01 22:16:30 by sof]
remove CCONST_xxx
sof [Wed, 1 May 2002 18:48:07 +0000 (18:48 +0000)]
[project @ 2002-05-01 18:48:07 by sof]
fix a silly introduced by prev. commit
sof [Wed, 1 May 2002 17:56:58 +0000 (17:56 +0000)]
[project @ 2002-05-01 17:56:52 by sof]
mapM ~> mapM_
sof [Wed, 1 May 2002 17:51:35 +0000 (17:51 +0000)]
[project @ 2002-05-01 17:51:35 by sof]
BUILD_LIB: also include EXTRA_OBJS
sof [Wed, 1 May 2002 17:51:02 +0000 (17:51 +0000)]
[project @ 2002-05-01 17:51:02 by sof]
define EXTRA_OBJS, derived from EXTRA_SRCS
sof [Wed, 1 May 2002 17:12:24 +0000 (17:12 +0000)]
[project @ 2002-05-01 17:12:24 by sof]
remove addDLL, now in InteractiveUI
simonmar [Wed, 1 May 2002 15:48:48 +0000 (15:48 +0000)]
[project @ 2002-05-01 15:48:48 by simonmar]
Clean up the code to load a dynamic library some more and hopefully
make it work on Windows (untested).
simonmar [Wed, 1 May 2002 15:46:15 +0000 (15:46 +0000)]
[project @ 2002-05-01 15:46:14 by simonmar]
Cleanup and overhaul the bogus dynamic library loading code in
InteractiveUI. Bugs fixed:
- when linking in extra_libraries from a package,
we now search library_paths in addition to the default
dlopen() paths.
- the path-searching machinery for dynamic libraries specified
on the command line was broken, it didn't work unless the library
was found on the first path in the list.
simonmar [Wed, 1 May 2002 13:16:04 +0000 (13:16 +0000)]
[project @ 2002-05-01 13:16:04 by simonmar]
Re-add reallyUnsafePtrEquality# (hey, it's really easy to add primops
these days!).
simonmar [Wed, 1 May 2002 13:12:11 +0000 (13:12 +0000)]
[project @ 2002-05-01 13:12:11 by simonmar]
Comment only
simonmar [Wed, 1 May 2002 13:11:24 +0000 (13:11 +0000)]
[project @ 2002-05-01 13:11:24 by simonmar]
Do 'make boot' then 'make all' as separate steps (fixes problem with
parallel make).
simonmar [Wed, 1 May 2002 10:45:58 +0000 (10:45 +0000)]
[project @ 2002-05-01 10:45:58 by simonmar]
Mention Haddock and libraries projects.
simonmar [Wed, 1 May 2002 09:30:06 +0000 (09:30 +0000)]
[project @ 2002-05-01 09:30:04 by simonmar]
- When converting ModuleNames to Modules for use in the the module
initialisation code, look them up in the IfaceTable(s) instead of
calling findModule again. They are guaranteed to be in either
the HomeIfaceTable or the PackageIfaceTable after the renamer,
so this saves some trips to the filesystem. Also, move this
code earlier in the compilation cycle to avoid holding on to the
renamed syntax for too long (not sure if this makes a difference or
not, but it definitely looked space-leakish before).
- remove Util.unJust, it is a duplicate of Maybes.expectJust
simonpj [Wed, 1 May 2002 09:10:32 +0000 (09:10 +0000)]
[project @ 2002-05-01 09:10:32 by simonpj]
When reading an interface file, don't discard CPR info even
if -fcpr-off is selected. The info describes what the .o file
contains, so we can't ignore that!
simonmar [Tue, 30 Apr 2002 16:34:12 +0000 (16:34 +0000)]
[project @ 2002-04-30 16:34:12 by simonmar]
Add a FAQ section. This will replace the mini-FAQ on the download
page (which we never really kept up to date), and will hopefully
accumulate all sorts of useful Q/As over time. Please feel free to
add to it...
simonmar [Tue, 30 Apr 2002 16:23:47 +0000 (16:23 +0000)]
[project @ 2002-04-30 16:23:47 by simonmar]
Remove some out of date stuff and tidy up the markup.
simonmar [Tue, 30 Apr 2002 15:34:24 +0000 (15:34 +0000)]
[project @ 2002-04-30 15:34:24 by simonmar]
Enclose the text about the CPP/string gap problem in a section so we
can refer to it easily.
simonmar [Tue, 30 Apr 2002 15:11:40 +0000 (15:11 +0000)]
[project @ 2002-04-30 15:11:40 by simonmar]
Add a section entitled "How do I tell if my bug has already been reported?"
simonmar [Tue, 30 Apr 2002 14:40:56 +0000 (14:40 +0000)]
[project @ 2002-04-30 14:40:56 by simonmar]
Make this work with 'make -j'
simonmar [Tue, 30 Apr 2002 14:40:33 +0000 (14:40 +0000)]
[project @ 2002-04-30 14:40:33 by simonmar]
hsc2hs isn't needed for 'make boot' any more (we do 'make boot all' in
fptools/ghc before booting fptools/libraries).
simonmar [Tue, 30 Apr 2002 14:19:56 +0000 (14:19 +0000)]
[project @ 2002-04-30 14:19:56 by simonmar]
Fix for 'make -j'
simonmar [Tue, 30 Apr 2002 10:31:50 +0000 (10:31 +0000)]
[project @ 2002-04-30 10:31:50 by simonmar]
Mention that you can use :load without any arguments.
simonmar [Tue, 30 Apr 2002 09:26:14 +0000 (09:26 +0000)]
[project @ 2002-04-30 09:26:14 by simonmar]
As Matt Chapman pointed out to me, the argument to allocatePinned() is
in words, not bytes. Fix the 8-byte alignment tweak (yet again).
panne [Mon, 29 Apr 2002 19:24:57 +0000 (19:24 +0000)]
[project @ 2002-04-29 19:24:57 by panne]
Guess what: SLIT => FSLIT :-}
panne [Mon, 29 Apr 2002 18:42:04 +0000 (18:42 +0000)]
[project @ 2002-04-29 18:42:03 by panne]
(F)SLIT fixes, continued...
panne [Mon, 29 Apr 2002 17:47:02 +0000 (17:47 +0000)]
[project @ 2002-04-29 17:47:02 by panne]
Sync with latest (F)SLIT changes
simonmar [Mon, 29 Apr 2002 16:17:55 +0000 (16:17 +0000)]
[project @ 2002-04-29 16:17:55 by simonmar]
- Fix bootstrapped compilation,
- Add the following RULE:
text "abc" ==> ptext SLIT("abc")
so most of the time there shouldn't be any need to use SLIT().
I'll go around and apply the opposite of the above RULE once
I've convinced myself that the RULE does what it should.
simonmar [Mon, 29 Apr 2002 15:41:59 +0000 (15:41 +0000)]
[project @ 2002-04-29 15:41:59 by simonmar]
Add haddock to $(AllProjects)
simonmar [Mon, 29 Apr 2002 14:04:11 +0000 (14:04 +0000)]
[project @ 2002-04-29 14:03:38 by simonmar]
FastString cleanup, stage 1.
The FastString type is no longer a mixture of hashed strings and
literal strings, it contains hashed strings only with O(1) comparison
(except for UnicodeStr, but that will also go away in due course). To
create a literal instance of FastString, use FSLIT("..").
By far the most common use of the old literal version of FastString
was in the pattern
ptext SLIT("...")
this combination still works, although it doesn't go via FastString
any more. The next stage will be to remove the need to use this
special combination at all, using a RULE.
To convert a FastString into an SDoc, now use 'ftext' instead of
'ptext'.
I've also removed all the FAST_STRING related macros from HsVersions.h
except for SLIT and FSLIT, just use the relevant functions from
FastString instead.
sof [Sun, 28 Apr 2002 01:58:29 +0000 (01:58 +0000)]
[project @ 2002-04-28 01:58:29 by sof]
Heal HEID.
* eRROR_ID: use FSLIT() rather than SLIT() to have "error" be added to
the FastString hash table rather than as an ext. C string; the Binary
instance for FastString can't handle the latter, only the former.
[any deep reasons why not?]
sof [Fri, 26 Apr 2002 22:35:55 +0000 (22:35 +0000)]
[project @ 2002-04-26 22:35:54 by sof]
When the OS thread is all out of work, shut down in a gentler
manner by calling shutdownHaskell() rather than
shutdownHaskellAndExit(). Give up the sched_mutex lock also.
sof [Fri, 26 Apr 2002 22:31:31 +0000 (22:31 +0000)]
[project @ 2002-04-26 22:31:31 by sof]
finalizeWeakPointersNow(): synchronously finalize, don't use sep. OS threads (if available.)
simonpj [Fri, 26 Apr 2002 16:32:04 +0000 (16:32 +0000)]
[project @ 2002-04-26 16:32:03 by simonpj]
Fix the bug that Sven found when bootstrapping:
stgSyn/CoreToStg.lhs:1112:
Couldn't match `#' against `*'
When matching types `GHC.Prim.Int#' and `a'
Expected type: GHC.Prim.Int#
Inferred type: a
In the application `error ("cafRefs " ++ (showSDoc (ppr id)))'
I forgot to keep eRROR_ID in the list of wiredInIds in MkId.
Fixed and commented (but not yet tested).
Simon
panne [Thu, 25 Apr 2002 18:48:02 +0000 (18:48 +0000)]
[project @ 2002-04-25 18:48:02 by panne]
Ooops, I mixed up PowerPC and SPARC ABIs: On SPARC the stack is always
double-word aligned (quad-word on PowerPC), so the "fully correct"
solution is very probably really correct. => Fixed the comment. :-}
panne [Thu, 25 Apr 2002 18:18:56 +0000 (18:18 +0000)]
[project @ 2002-04-25 18:18:56 by panne]
Highly untested fixes for SPARC:
* Commented on "fully correct" solution
* Flush cache
* Fixed freeHaskellFunctionPtr
FYI: http://www.cs.unm.edu/~maccabe/classes/341/labman/node12.html
contains a good "Dummy's Guide" to the SPARC stack frame layout.
Just if you can't dig this info up on e.g. SUN's site, because
they've hidden everything useful... :-]
sof [Thu, 25 Apr 2002 15:42:46 +0000 (15:42 +0000)]
[project @ 2002-04-25 15:42:46 by sof]
add thenIO to knownKeyNames - as was, ghci-HEAD fell over
simonmar [Thu, 25 Apr 2002 08:57:51 +0000 (08:57 +0000)]
[project @ 2002-04-25 08:57:51 by simonmar]
Remove a bogus ASSERT (I *knew* I should have tested that
last change properly... one bug per line, sigh).
sof [Thu, 25 Apr 2002 04:54:55 +0000 (04:54 +0000)]
[project @ 2002-04-25 04:54:55 by sof]
allocatePinned: do what the prev. commit was billed to do
ken [Thu, 25 Apr 2002 00:29:59 +0000 (00:29 +0000)]
[project @ 2002-04-25 00:29:59 by ken]
-lpthread requires -lexc on alpha-dec-osf3
simonmar [Wed, 24 Apr 2002 13:29:01 +0000 (13:29 +0000)]
[project @ 2002-04-24 13:29:01 by simonmar]
freeStablePtr(): Only call freeStableName() if sn->ref == 0.
From: Bernard James POPE <bjpop@cs.mu.OZ.AU>
simonpj [Wed, 24 Apr 2002 11:06:11 +0000 (11:06 +0000)]
[project @ 2002-04-24 11:06:10 by simonpj]
Fix an obscure corner situation in case-simplification (cg051)
simonpj [Wed, 24 Apr 2002 10:12:53 +0000 (10:12 +0000)]
[project @ 2002-04-24 10:12:52 by simonpj]
Do the scoped-tyvar fix right this time
simonmar [Wed, 24 Apr 2002 09:42:18 +0000 (09:42 +0000)]
[project @ 2002-04-24 09:42:18 by simonmar]
There's no gmp package these days
ken [Tue, 23 Apr 2002 17:33:54 +0000 (17:33 +0000)]
[project @ 2002-04-23 17:33:54 by ken]
Replace strncmp() with memcmp(), just to be safe
ken [Tue, 23 Apr 2002 17:16:01 +0000 (17:16 +0000)]
[project @ 2002-04-23 17:16:01 by ken]
#define _POSIX_C_SOURCE 199506L, so that we get pthread stuff properly on
alpha-dec-osf3.
simonmar [Tue, 23 Apr 2002 15:49:43 +0000 (15:49 +0000)]
[project @ 2002-04-23 15:49:43 by simonmar]
PrelPrimopWrappers -> GHC.PrimopWrappers (in debugging output)
simonmar [Tue, 23 Apr 2002 15:49:22 +0000 (15:49 +0000)]
[project @ 2002-04-23 15:49:22 by simonmar]
Urk, PrelPrimopWrappers is now GHCziPrimopWrappers (sigh, this should
really be done in a less fragile way).
simonmar [Tue, 23 Apr 2002 15:41:58 +0000 (15:41 +0000)]
[project @ 2002-04-23 15:41:58 by simonmar]
Add some folklore about Perl.
sof [Tue, 23 Apr 2002 14:20:18 +0000 (14:20 +0000)]
[project @ 2002-04-23 14:20:18 by sof]
forkProcess() mingw unused arg tweak
simonmar [Tue, 23 Apr 2002 11:22:12 +0000 (11:22 +0000)]
[project @ 2002-04-23 11:22:12 by simonmar]
Ensure that the *goods* in the array returned by newPinnedByteArray#
are always 8-byte aligned.
simonmar [Tue, 23 Apr 2002 11:21:39 +0000 (11:21 +0000)]
[project @ 2002-04-23 11:21:39 by simonmar]
Always return 8-byte aligned memory from allocatePinned().
simonmar [Tue, 23 Apr 2002 11:11:15 +0000 (11:11 +0000)]
[project @ 2002-04-23 11:11:15 by simonmar]
Remove unnecessary \ at the end of lines
stolz [Tue, 23 Apr 2002 09:56:29 +0000 (09:56 +0000)]
[project @ 2002-04-23 09:56:28 by stolz]
- set thread labels to NULL after free()
- labelThread# didn't exit properly
TODO: Labels are prematurely free()d when a thread terminates. Better let
the GC worry?
simonpj [Tue, 23 Apr 2002 08:55:47 +0000 (08:55 +0000)]
[project @ 2002-04-23 08:55:47 by simonpj]
Reject attempts to rebind built in syntax
simonpj [Tue, 23 Apr 2002 08:52:16 +0000 (08:52 +0000)]
[project @ 2002-04-23 08:52:16 by simonpj]
Without -fglasow-exts, quantify over all mentioned type vars in type sigs
simonmar [Tue, 23 Apr 2002 08:40:52 +0000 (08:40 +0000)]
[project @ 2002-04-23 08:40:52 by simonmar]
Need to pull in thenIO when compiling a stmt now.
sof [Tue, 23 Apr 2002 06:34:27 +0000 (06:34 +0000)]
[project @ 2002-04-23 06:34:26 by sof]
More sched_mutex related cleanup & fixes; little bit
too delicate for my liking, this.
* Revert lock assumptions for raiseAsync(); now assumed
to run with sched_mutex held (as it mucks about with a
TSO).
* stodgily release / acquire sched_mutex around calls
to startSignalHandlers() (as is done in Signals.c)
* in the presence of user-installed signal handlers, the
MT-enabled RTS failed to shutdown -- all queues empty
causes a lone RTS worker thread to pause() waiting for
signals. This is not the right thing to do; I (temporarily?)
disabled this signal-wait fallback in MT mode and shut
down instead. We need to be clearer as to what is a shutdown
condition in MT mode.
* The use of sched_mutex to protect next_thread_id increments
is causing headaches; disabled in non-SMP mode right now until
I've figured out the pthreads-correct way of doing atomic
increments.
* There's still a ^C-related problem which causes the Haskell
handler to sometimes induce a SEGV when run. Feel free to debug :)
sof [Tue, 23 Apr 2002 06:19:06 +0000 (06:19 +0000)]
[project @ 2002-04-23 06:19:06 by sof]
Simplified yieldToReturningWorker(), no need to have OS thread yield