GitLab

wibble

Drop the dependency on the Win32 library, as it is only used in an
extremely limited way (a single 'foreign import' is all that's reqd)
[Sorry Reuben, I don't buy your 'extra deps are good' argument :-) ]

runSomething: narrow down prev commit, only quote args under mingw;it interacts badly with -D quoting already done on other plats (and isn't reqd)

Add comment and use backslash consistently

Remove trace

--------------------------------
	First cut at the demand analyser
	--------------------------------

This demand analyser is intended to replace the strictness/absence
analyser, and the CPR analyser.

This commit adds it to the compiler, but in an entirely non-invasive
way.

	If you build the compiler without -DDEBUG,
	you won't get it at all.

	If you build the compiler with -DDEBUG,
	you'll get the demand analyser, but the existing
	strictness analyser etc are still there.  All the
	demand analyser does is to compare its output with
	the existing stuff and report differences.

There's no cross-module stuff for demand info yet.

The strictness/demand info is put the IdInfo as
	newStrictnessInfo
	newDemandInfo

Eventually we'll remove the old ones.

Simon

Reset the terminal on exit so that ghci doesn't leave it messed up. To see
what I mean, run ghci, then type :q RET, then su return, and start typing a
password. Alarming, isn't it?

This fix needs a patch to the Readline library (next commit).

Make -u's and symbols into separate args

1. Make GHC work with paths and filenames containing spaces (hopefully).
This is done by quoting all arguments sent to the shell. To avoid causing
problems, this means that each argument in a list had *better* be a single
argument, hence the next commit to PackageSrc.hs.

2. undosify the path passed as the -h argument to unlit, to make unlit work
on Windows again.

3. Fix getExecDir for Windows (broken by previous changes to the meaning of
-B).

Typos in a comment. Whitespace at eols.

Fix comment.

Long long functions are no longer in the rts.

w

Follow the s/includes/include/ change in package.conf.

Add ghc-usage.txt to install tree info

GHC_INCLUDES => GHC_INCLUDE_DIR

Modify Windows installation instructions in accordance with the Way Of
The Future.

Wibbles

Fix markup bugs; add documentation about deprecations

Add docs re installed file layout

Remove spurious space.

Don't add -I flag for Windows gcc's include files here; do it in
package.conf instead.

Add C includes path for Windows here rather than hardwiring into
SysTools.lhs. Change to include rather than includes.

Now assume that -B passes libdir, not libexecdir

Binaries now live in extra-bin

Fix binary finding for Unix

Tidy up Type/TcType stuff in DsCCall/DsForeign

Set -DDLLized when DLLized is true for use in C source

Add stubs for ILXized cbits DLL.

Don't call resetNonBlockingFd when DLLized; this stops the cbits DLL
depending on the RTS, which ILX needs; it doesn't matter for Windows, as
that doesn't have non-blocking IO implemented anyway.

Fix DLL building, and add support for linking .hsc.o files into cbits libraries

s/extra-bin/bin/ because this is where all programs are installed to.
Perhaps it should be changed the other way around - I don't know what
this extra-bin was supposed to be.

Fix paths in installed hsc2hs driver.

hscStmt: track changes made to the TC entry points typecheckExpr, typecheckStmt

eqForeignPtr: implement equality using new primitive

Heal HEID

- eqForeignObjZh in include/PrimOps.h didn't have quite the right
  shape (the result is a macro arg). hslibs/lang/ForeignObj
  wasn't up on the change to eqForeignObj now being a primop.

- recent ghc/compiler/deSugar/ changes broke the handling of
  CCall & FFI decls quite a bit. Backed out most the rewrites
  of Type.splitFoo to TcType.tcSplitFoo (i.e., now back to using
  TcType.tcSplitFoo).

  The backed-out newtype-related changes were by no means accidental.
  But, I don't profess to understand their intention to make the
  proper fix, so my change is just a stop-gap measure to get HEAD
  back to the land of the living.

- recent changes to the behaviour of 'hiding' & qualified names
  broke hslibs/lang/CString hslibs/data/edison/Seq/ListSeq,
  hslibs/data/edison/Coll/TestOrdBag, hslibs/data/edison/Coll/UnbalancedSet,
  hslibs/data/edison/Coll/TestOrdSet, hslibs/data/edison/Seq/TestSeq

- rename 64-bit 'primop' funs that now live in lib/std/cbits/longlong.c
  back to what they used to be called (i.e., prefixed with "stg_").
  Why?
     - less likely they'll clash with other (user supplied) entry points
       at link-time.
     - matches protos in ghc/includes/PrimOp.h

add warning about not changing WhatGccIsCalled from build.mk.

remove superfluous "return ()"

error msg wibble

check TMPDIR (got lost somewhere in the conversion to SysTools)

Don't build RTS for ILX