GitLab

No signal support with mingw32

Win32 platforms: use native API to compute elapsed and user times.

When putting the RTS in a DLL, we have to delay filling in the
Charlike and Intlike tables until load-time.

StgNat64 --> StgWord64

typo

StgNat64 --> StgWord64

initStorage: initialise gos0->to_space

- misc changes to support DLLs
- StgNat* --> StgWord*

Partial support for working with DLLs. On mingw32, unless you supply
the -static command line option, the default is now to assume
that you're using DLLs when compiling and linking.

ToDo: support the construction of DLLs via the driver.

On mingw32, which is the only 'platform' where we support producing
DLLs, prefix each static closure with a zero word. This is needed so
that we can distinguish between pointers to (reversed!) info tables
and static closures just by checking whether there's a zero word just
above the pointed-to entity. Wish there was a better way..

Remove out-of-date use of -fcompiling-prelude

Compile CgCase with -fno-prune-tydecls on

import list adjustments

One new option:

 -static    produce code for use in statically linked binaries.
            The driver makes sure that this option is on for
	    platforms that doesn't support DLLs. Leaving it
	    off won't have any bad effects though.

ppr bugfix for dcons that are locally qualified
(omit braces around context if there isn't one.)

Directories can now be flagged as containing interface files that have
their corresponding object codes living in Win32 DLLs.

The compiler needs to keep track of whether a name refers to something
in a DLL or not, since Win32 DLLs forces you to distinguish between
the two at the point of use. For example, the code generated for
the following snippet

     return (x+2);

will differ. If 'x' resides in a DLL, you need to perform an extra
indirection to get at its value. Effectively, the generated code
becomes

     return (*x+2);

For functions, the distinction can be made transparent, but we
can avoid jumping through an extra level of indirection if we
do indicate that a label will be imported from a DLL.

Back to the renamer and its scheme, directories that contain
the file ".dLL_ifs.hi" (name chosen to lessen the risk of a clash..)
are considered as containing 'DLL interface files'. There's two
caveats to this scheme:

 - interface files found in "." are not considered to be referring
   to something in a DLL.
 - if the compiler has got -static on the command line, then all
   interface file in scope are considered to be 'normal'.

Win32 only: emit code that declares the DLLness of a label we're
making use of.

Add missing default case to mkRegLiveness.

Support for deriving 'proper' Show & Read instances for infix constructors.

PrelMods - use mkPrelModule instead of mkSrcModule to reflect distinction
we need to make between the two when compiling up code that's destined
for 'Win32 DLL'dom.

Fix to allow local, non-exported actions to be 'foreign export'ed.

- import list tweaks
- moved the code that decides that a StgCon really shouldn't
  be mapped to a static constructor but an updateable thunk
  if it contains lit-lits from the codegen into the CoreToStg
  translation.

  Added an extra case to this code to deal with StgCon's that contain
  references to values that reside in a DLL, where we also have to
  opt for an updateable thunk instead of a static constructor. Only
  applies when compiling on/for Win32 platforms.

mostly import list re-shuffling

Separated out Module type and ops from OccName.

Install the docs in $real_datadir, not $datadir.

Tweaks:
  - only trailing empty decls were accepted.
  - empty export lists where being flagged as there being none.

Updates + added empty decl regression test

makeForeignObj --> mkForeignObj

hppa1.1 support, first cut.

recurse into gmp on 'make clean'

Some native codegen updates.

hppa1.1 support, first cut.

Use 'data_start' as the end-of-text marker if it exists.  This includes
the read-only text section on Linux/ELF, which is what we want.

Bring in rev. 1.22.2.1 - Don't apply -monly-n-reg patches to _regBlah
functions (profiling cost-centre register things)

Minor Makefile mods.  Add a Prelude.hs which suits new STGhugs.

Mods to make STG-hugs able to compile and run small examples.  This
commit also includes proper implementations of seq, raise and catch.

Missing { in mulIntCzh

Handle block comments that haven't been closed a little bit better: record
(and report) the start of the comment.

Simple implementation - doesn't bother dealing with nested comments.

typo

import list tweak