GitLab

Avoid loops in CCS graph.

Don't wake up threads if the select() was interrupted by a signal.

Removed obsolete comment.

Changed __init_Main to __init_PrelMain in line with changes elsewhere.

Call __init_PrelMain instead of __initMain.  PrelMain lives above Main
in the dependency tree, so this is more correct.

Added PrelException_BlockedOnDeadMVar_static_closure and
PrelException_NonTermination_static_closure.

Made them build on mingw32 again (problems with getourtimeofday, which
shouldn't be needed under mingw anyway).

ITimer.h is really called Itimer.h. Hmmm, let's guess on what
brilliant OS this part of the RTS was developed...  :-))

Adding an alternative to the "delay" system used for
threads that are waiting for time to pass.

This works on a target time basis, eliminating the
need to use the ticky style counter.

It is only enabled under:

#if defined(INTERPRETER) && !defined(HAVE_SETITIMER)

Second attack at supporting threads inside STG Hugs.
We now support most of the concurrency primitives.

Also a wibble in Evaluator.c, letting Hugs compile.

Added PrelException_PutFullMVar_static_closure and __init_Prelude, and also
(rather hackily) __init_Main (until I find out how that should really be
dealt with).

More random cleanups.

Random cleanup: rename the TSO->whatNext field to what_next, to be a
bit more consistent.

Support "interruptible operations": threads which are blocked on
MVars, exceptions, I/O, or sleeping are fair game for raiseInThread.

bug fix in previous commit - NotBlocked is a valid state to be in in
resurrectThreads().

Raise an exception in any thread that is about to be garbage
collected.

	- threads blocked on MVars that are otherwise garbage
	  will receive a BlockedOnDeadMVar exception.

	- threads blocked on Black Holes must be in some kind
	  of deadlock, so we send them a NonTermination exception.

put a lock in the right place.

decruftification

Use Sp instead of the global init_stack for the stack pointer during
module initialisation.  This should cut down the code size for module
initialisation blocks.

Unbreak signals.

Fix configuration/building of GMP.

Update Hugs so as to work with the new way of referencing Prelude
symbols from the RTS.  The main action is in fixupRTStoPreludeRefs()
in rts/Prelude.c.

New file for RTS->Prelude dependencies.

Handle references from the RTS to the Prelude in a more consistent
way.

	- For statically-linked binaries, nothing has changed.

	- For the interpreter, refs from the RTS to the Prelude
	  are now indirected.  The indirections need to be
	  filled in at some point during startup by calling
	  fixupPreludeRefs (in Prelude.c).

	- The CHARLIKE and INTLIKE tables are now handled in
	  the same way for both Hugs and DLLs.

Hugs will be broken for a short while until Julian sorts out the Hugs
parts of this change.

Don't be selective about what's exported from the RTS DLL any more,
as it requires sync'ing HSrts.def whenever the RTS offerings change.

* 'inlined' subset of the MM protos needed to compile this module, i.e.,
  don't require whoever does a 'make' in ghc/rts/ to have extended their
  <windows.h>
* right Win32 proto for handle_tick()

Support for setting target-specific CC options via AC_SUBST (and make use
of it on the mingw32 front -- 'resist' the temptation to use GMP's config/
story for this.)

mingw32 support

remove some Makefile debugging bits

When booting, pass in --target to gmp/

Track recent re-homing of Prelude.h for INTERPRETER.

Move Prelude.h from ghc/includes into ghc/rts.  It's essentially a
list of the things from the Prelude that the RTS depends on, and isn't
relevant to STG code.  Furthermore, this helps to clarify which parts
of the RTS depend on things from the Prelude (by an explicity
#include).

Clean up a couple of compilation warnings.

Remove #ifdef STANDALONE_INTEGER.

Remove duplicate definition of unblockOne.

Reset the interrupted flag after an interruption.  This allows
finalizers to run, and means that Hugs doesn't get stuck in an
interruption loop.

Disable module initialisation under Hugs until we can sort out the
hairy linking issues.

- generalise the per-module initialisation stubs so that we use it
  in normal (non-profiled) code too.  The initialisation stubs are
  now called '__init_<module>' rather than '_reg<module>'.

- Register foreign exported functions as stable pointers in the
  initialisation code for the module.  This fixes the foreign export
  problems reported by several people.

- remove the concept of "module groups" from the profiling subsystem.

- change the profiling semantics slightly; it should be unnecessary
  to use '-caf-all' to get reasonable profiles now.

Attempt to fix the Sparc version of StgRun.  Since
RESERVED_STACK_SPACE increased beyond the size of a 13-bit immediate,
the Sparc code has been broken.

We try to fix this with some imaginative gcc inline assembly
constraints.  If this breaks again, I'll try recoding it directly in
assembler.

While I'm here, pull in some of the comments from the old pre-4.00 RTS
and add some new comments.

oops, need to define prof_file when DEBUGging too.