GitLab

Remove unbalanced #endif.

Use Sleep under mingw rather than sleep (which is obsolete), and try using
fork under cygwin (it should work by now, shurely...).

Don't declare unused struct termios on mingwin.

Move if #ifndefs so that no code is compiled on win32; the code that was
compiled was doing nothing useful, and probably shouldn't have been.

Add -mno-cygwin to CPPFLAGS on i386-unknown-mingw32 to stop configure.h
getting lots of bogus headers that aren't there. I'm not at all sure why
this hasn't caused problems before...

fix the GHC_HAPPY_OPTS hack to correctly detect happy 1.10

Rather than implement -K (which would probably need another primop),
simply bump the maximum stack size to 8M.  It dynamically grows
anyhow.

Expand on comments re StString lifting activity

Fix two bugs exposed when trying to boot HEAD on sparc with NCG and -O:

1.  StScratchWords on sparc were in the wrong place; they were
    immediately above %fp and should have been immediately below.
    Fixed.  Also removed a suspicious-looking "+1" in the x86
    version of same.

2.  (Potentially affects all platforms): Lift strings out from
    top-level literal data, and place them at the end of the block.
    The motivating example (bug) was:

     Stix:
        (DataSegment)
        Bogon.ping_closure :
        (Data P_ Addr.A#_static_info)
        (Data StgAddr (Str `alalal'))
        (Data P_ (0))
     results in:
        .data
                .align 8
        .global Bogon_ping_closure
        Bogon_ping_closure:
                .long   Addr_Azh_static_info
                .long   .Ln1a8
        .Ln1a8:
                .byte   0x61
                .byte   0x6C
                .byte   0x61
                .byte   0x6C
                .byte   0x61
                .byte   0x6C
                .byte   0x00
                .long   0
   ie, the Str is planted in-line, when what we really meant was to place
   a _reference_ to the string there.  This is Way Wrong (tm).  Fixed.

Oops, I broke it yesterday; don't use autoheader.
(I don't see my yesterday's commit log. I changed #define to #undef
in autoconf templates.)

Change #define to #undef in autoconf templates.

Recent Unicode and future ISO-10646 finally decided that the character
code space ends at U+10FFFF. Let ghc follow the rules: maxBound::Char
is now '\x10FFFF', etc.

The LIBM variable was between two Readline related variables. Cleaned up.

Make LIBM be detected by autoconf.

Today's doc hacking.

Remove ghc-$(ProjectVersion) from GhcBinDistPrlScripts, since it's not
a perl script any more.

Remove ghc-stats from GhcBinDistLibPrlScripts.

deLam the contents of an _scc_: they don't disappear at code
generation time.

Use ${1+"$@"} instead of "$@". How stupid are those shells.

Use "$@" instead of $* to pass commandline arguments safely.

More documentation rewriting...  I'm particularly proud of the "flag
reference" section, please check it out.

Correct usage info for tryPutMVar#.

VoidRep call/return support for interpreted code.

VoidRep call/return support for the interpreter.

re-order the WithHc and WithGhc tests: now if you specify
--with-ghc=<ghc>, then WithHc also gets <ghc> if you don't set it
explicitly.  Previously it got whatever ghc was in the PATH, which
would fail if there wasn't a ghc in the PATH at all.

changes to support tryPutMVar which I forgot in the main commit.

Remove -l from library name in libraries list (oops!)

simple tryPutMVar test.

Tidy the type we get back from typecheckExpr.

add tryPutMVarzh_fast

- make putMVar block rather than raise an exception when it encounters a
  full MVar (to match the semantics in our recent paper on async excpetions).

- add tryPutMVar, a non-blocking version of putMVar.

Turn off linking under Windows until it's fixed (that's linking, there's no
hope of fixing Windows).

Add winmm to libraries for RTS under Windows, where it's now used for
threadDelay.

Remove inadvertently committed debugging gumph.

Make isDirectory work on Windows (dirent.d_name is a char *, not a char []);
also getDirectoryContents (readdir helpfully sets errno to ENOENT when we've
read the last entry, rather than leaving it at 0).

Windows is *not* the way of the future.

Remove an out-of-date comment and correct a Windows wibble.

PrimOp dependency wibble

Remove support for directly loading .so's/.DLL's from the cmd line
in interactive mode and instead support the -lfooble mechanism.

Show error messages generated from failed .so/.DLL load attempts.

Allow addDLL to return error message strings to caller.