GitLab

----------------------------------
	Fix an existential-constructor bug
	----------------------------------

		DON'T MERGE!

This fixes a bug I introduced in my
tidying up for scoped type variables.

You'll get bizarre behaviour from the HEAD if
you use existentials, until you add this fix!

Clarify the situation with GHCi object files, and add instructions for
building one from a .a file.

markup wibble

markup fixes

Clarify where we look for libraries named on the command line (-lfoo),
and mention LD_LIBRARY_PATH.

The implementation is still wrong, however.

Add -mwin32 -mno-cygwin to CPPFLAGS.

Give correct output for mingw.

Initial mods to make the Glorious New IO Library (tm) work on mingw.
Not everything works, but is compilable, and off to a good start.

Fix a difference between the NCG implementation of the UPD_CAF() macro
and the C version.  I think the C version was changed recently but the
NCG version never caught up.  Fixes a crash in GHCi with NCG compiled
code.

Correct the instructions for compiling a debugging RTS (-optc isn't
needed before options in GhcRtsCcOpts any more).

missing whitespaces caused touch/rm to fail

Stuff about scoped type variables

Don't build lndir on mingwin, which doesn't have symlinks.

Reimplement getCPUTime in Haskell for Windows.

Deal with nested comments (HDirect sources showed up this bug)

Fix removing temporary files.

When we auto-load a module because the user typed a qualified name at
the prompt, we better not auto-load a home interface (because we won't
have the code to go with it).

So, introduce a new constructor in the WhereFrom datatype, namely
ImportByCmdLine for these auto-imports, and make findAndReadIface fail
if it tries to load a home interface by this route.

ToDo: GHCi should *never* demand-load a home interface under any
circumstances, but we don't have an ASSERT for this yet.

Change the GHCi monad from type

	GHCiState -> IO (GHCiState, a)
to
	IORef GHCiState -> IO a

to avoid losing recent changes to the state when we receive an
exception (which would normally be harmless, except that the state
isn't purely functional: it must match some state kept by the RTS's
dynamic linker).  Asynchonous exceptions could still cause us some
difficulty.

Wibble for scoped type variables

get rid of some more cpp warnings

de-ramble previous commit

Misc minor changes to integrate GHC a little bit better on Win32 platforms.
Specifically, the commit does the following (assuming you've configured
fptools/ with the option --enable-minimal-unix-deps on a mingw platform):

 * when GHC uses System.system, it expects an MSDOS command processor to
   interpret the command. This implies that 'normal' UNIX shell utils will
   no longer be used, but substituted with MSDOS equivalents.
 * the GHC backend relies on gcc and perl to handle .s/.hc/.o/.a files. GHC
   will now assume that these all live in one 'tool directory', making it
   easier to bundle these backend tools with GHC.

The upshot of these changes is that it is now possible for the user not to
have to install cygwin prior to installing GHC (as the upcoming ghc-win32
binary release will prove).

Avoid the use of Posix when compiling for target i386-unknown-mingw32

Added GHC_TOUCHY_DIR and GHC_TOUCHY

New configure script option, --enable-minimal-unix-deps, a Win32-specific
option which lets you build a compiler that doesn't depend on cygwin tools
to run. Well, you still need tools such as gcc and perl in the backend, but
these don't have to be cygwin ports.

wibble

'flex' or 'lex' is required to compile up fptools/ (e.g., reqd in
glafp-utils/sgmlverb). autoconf's standard macro AC_PROG_LEX just
defaults to 'lex' if 'flex' isn't found, but doesn't actually
check whether 'lex' is available along PATH.

Added AC_PROG_LEX_STRICT, which extends AC_PROG_LEX by checking
for the presence of 'lex' if 'flex' isn't found. It bails
out with an error message if 'lex' cannot be found.

wibble - silence CPP warning

wibble - silence CPP warnings

wibble - drop a ? from a fprintf format string to avoid having it look like a trigraph seq might be present

Minor tweaks to IlxGen backend

Check for re_search_2() when checking whether libc has GNU regex.
This is a temporary fix until we get rid of GNU regex altogether and
do something more portable.

*** MAYBE MERGE ***

NetBSD is ELF only now.

*** MAYBE MERGE (and the last revision) ***

we're only supporting NetBSD/ELF now, not a.out (Lennart's suggestion)

wibble

Add -lm to HC_BOOT_LIBS

*** MAYBE MERGE ***

Link with HStext_cbits in case we need regex (bug from Lennart).

*** MAYBE MERGE ***

netbsd_aout => netbsdaout

NetBSD <= 1.4 is a.out, apparently

NetBSD is ELF by default.