GitLab

Fix eta-expansion bug, which led to an infinite loop in
CoreUtils.etaExpand. This showed up when compiling FranTk.

		MERGE TO STABLE

Add notes about DEBUG and GHCI.

update (readline has moved)

Add a ToDo note.

Further mangler changes to get ghc working with gcc 3.04 on the Alpha.
Jeff Lewis: "The compiler was sometimes emitting the $..ng label for a symbol
before the regular label.  This really confused the mangler, and it completely
scrambled the file."

Add the beginnings of the "Coding Style Guidelines" for ghc/compiler.

Remove duplicate sched_belch() call from raiseAsync().

Add notes about data types

Fix braino in previous commit

MERGE TO STABLE

- If the same module is defined in multiple "root" source files, then
  complain.  Before, the compiler would silently ignore one of them.

- Remove some unuseed imports

MERGE TO STABLE

Fix a bug in the cleaning up of temporary files: we were wrongly
cleaning the temporary pre-processed source files for stable modules,
with the result that the compiler would panic the next time they were
required.

Reported ages ago by: Hal Duame.

MERGE TO STABLE

Replace "Funny global thing" with a better error message, as suggested
by Alastair Reid.  The message is:

  Warning: retaining unknown function `$_' in output from C compiler

("unknown function" is a bit vague, but I couldn't come up with an
alternative that wasn't misleading).

bring write() proto into scope

Use __DISCARD__ to prevent overly aggressive optimization by certain
gcc versions on the Alpha.  Thanks to Jeffrey Lewis!

Add link to Mark Jones' paper on functional dependencies, in lieu of
real documentation.

fromInt ==> fromIntegral

Oops, string literals were using mkFastStringNarrow instead of
mkFastString, with the result that characters >= 256 were being
truncated modulo 256.

MERGE TO STABLE

Fix a buglet in CorePrep:  an expression such as

	case __coerce (\x -> e) :: T of { ... }

would be left as is, but the lambda expression should really be
abstracted as a let (causes a panic later in srtExpr; shown up by
the dynamic001 test).  There was a missing call to deLam in the case
for Case expressions in corePrepExprFloat.

In addition, I made a new version of deLam, deLamFloat, which can
float any bindings generated.  This helps to generate slightly cleaner
code in the above case (the binding is floated out of the scrutinee).

Also: GC unused imports while I'm here.

Protect against divide by zero when there have been zero major GCs in
the one-line stats mode.

oops, rts_n_free_capabilities is used in an inline function.

Use an stg_ap_2 thunk rather than an AP_UPD in rts_apply().  Saves one
word per rts_apply(), and a little time too.

Global and common variable sweep: staticize many variables that don't
need to be globally visible.

Re-order the checks in HP_STK_CHK and HP_STK_CHK_NP: if the stack
check failed before the heap check, then we would get bogus allocation
results.

large_alloc_list seems to be a fossil; remove it

Be a bit clearer about the current directory when GHCi is started, and
mention that you might have to ':cd' to the right place before you can
load something.

If booting unregisterised, add -DNO_REGS -DUSE_MINIINTERPRETER
to PLATFORM_HC_BOOT_CC_OPTS

Update docs about not using Float.

HC_BOOT_LD_OPTS:
  - if needed, include ghc/rts/gmp on -L list.
  - sync -u list wrt ghc/rts/rts.conf.in

oops, YES is NO

UseGhcForCc: don't use "ifneq" to control its setting; too strict.
Resort to a $(shell ..) hack.

Added GHC_MANGLER_DIR, GHC_SPLIT_DIR + fix .raw_s -> .s bootstrap suffix rule

NON_HS_PKG vs NON_HS_PACKAGE confusion - NON_HS_PACKAGE wins

Make this compile, and maybe even work

- Update the old coding style document
- Move it to the commentary under the "RTS & libs" section

Overhaul the debugging flags.  The old bitfield mechanism was a bit
silly, so now instead of writing -D129 (the one I use most) you can
now write -DSs.  Each debugging option has a single character assigned
to it, documented in the -? help text.

We have to turn off the buffering on stdout, stdin & stderr after
every time we revert all CAFs, not just when starting up GHCi.

MERGE TO STABLE

PrimRep Cleanup

   - Remove all PrimReps which were just different flavours of
     PtrRep.  Now, everything which is a pointer to a closure of
     some kind is always a PtrRep.

   - Three of the deleted PrimReps, namely ArrayRep, ByteArrayRep,
     and ForeignObj rep, had a subtle reason for their existence:
     the abstract C pretty-printer(!) used them to decide whether
     to apply a shim to an outgoing C-call argument: a ByteArrayRep
     argument would be adjusted to point past the object header,
     for example.

     I've changed this to happen in a much more reasonable and
     obvious way: there are now explict macros in AbsCSyn to do the
     adjustment, and the code generator makes calls to these as
     necessary.  Slightly less hackery is necessary in the NCG as
     a result.

Pass -fno-builtin to gcc when compiling .hc files, because this avoids
a common cause of running out of registers when foreign calls to
functions like memcpy(), strlen() are made from Haskell.

MERGE

Remove out-of-date comment

Mention that --show-iface is needed for viewing interface files.