- User Since
- Oct 5 2016, 6:50 AM (137 w, 2 d)
Dec 15 2017
Fix unused variable error
Dec 13 2017
Pass unsafeGlobalDynFlags where it is needed
Fix compilation errors
Oct 13 2017
Is there a way we can quit using GCC on Windows and just call the linker
Apr 9 2017
- Merge branch 'master' into computed-goto
Mar 19 2017
This seems horribly inefficient. It seems that the best approach would be something integrated into the compiler backend. Especially on x64 where the SSE registers are used.
Feb 15 2017
Feb 2 2017
Make M4 quoting more consistent
Some Autoconf fixes
Feb 1 2017
Pass -Werror when checking if the C compiler supports a flag
Try to fix macOS build
Jan 31 2017
Fix configure. Don't goto from do-while loop.
Jan 29 2017
Clean up CPP in rts/Interpreter.c
Another aclocal.m4 fix
Jan 28 2017
Fix configure syntax error
Autoconf checks for flags for fast computed goto
Jan 26 2017
Jan 25 2017
Jan 24 2017
Hopefully fix the segfault for real.
Disable computed gotos in the DEBUG RTS
Jan 23 2017
Don't clobber an old recipe for includes/rts/Bytecodes.h
Jan 22 2017
@ezyang no, I just noticed this is a very old active revision.
Jan 21 2017
Fix another typo
Add missing makefile
Typo in comment
Revert some makefile changes
@bgamari How can I make the build system work? I am stuck.
Try to unbreak the build
Fix build failure
Fix haddock submodule
Fix haddock submodule
Fix compilation error
Jan 20 2017
put more opcodes on one line
Fix an assertion and comment
Fix INTERP_STATS build
Jan 19 2017
Jan 18 2017
@simonmar LuaJIT has a finely tuned linear scan allocator, which might be a
source of information.
Jan 17 2017
Revert changes to other shell code; this will go in other patch
Forgot about this revision. It is superseded by D2986 anyway, so I'm abandoning it.
Jan 16 2017
Do you plan on fixing the graph allocator at some point?
@bgamari I agree (and that is exactly what I ended up doing).
Ah. So in this case converting to SSA form would solve the problem.
Jan 15 2017
Looks good to me. Works with LLVM version 3.8 too.
Jan 14 2017
@erikd The ghci-ext-prof wrong exit codes are in the stdout output of the build.
I think that this is good, but I am somewhat worried about the T12234 stat failure and the failures in the ghci-ext-prof ways.
@shlevy How does one do that? Would it pass App Store approval?
The runtime linker can't work on iOS, to the best of my understanding. I believe iOS doesn't allow apps to mark code as executable at runtime. libFFI uses statically mmap'd trampolines to work around this problem.
I don't like this. mk/build.mk is meant to be customized by the user, and so shouldn't be deleted.
Jan 9 2017
Update user's guide for HeapOverflow being throw
Jan 7 2017
Fix documentation of getTopHandlerThread
Accept the new T5644 output
Fix T5644 to account for the running of finalizers on heap overflow
Jan 3 2017
Add comments and validity check
Revert back to working version
- Enable subsections via symbols on iOS, too
Jan 2 2017
Revert a change that affected non-Darwin platforms
Use -dead_strip only when linking a final object file
Fix unbound variable error
Don't use -Wl,-r on Darwin – it is incompatible with -dead_strip
Jan 1 2017
Import what is needed from Platform
Fix silly type error
Dec 26 2016
Dec 22 2016
Fix units of heapLimitGrace
Delete unnecessary #include
Add a test for catching HeapOverflow
Dec 21 2016
Clarify the help for +RTS -Mgrace=