GitLab

Mention GHC's relaxation of the layout rule

* set off examples (e.g. as used in the Cabal doc) from the text.

* make minor headers blue too.

* make the green in screens a bit darker for contrast.

wibble to previous... (*fatalInternalErrorFn)() is not expected to return.

assertion failures should go through the RtsMessages layer, so they
get a pop-up box in a Windows app.

Recommend using -M rather than -H to increase the heap size.

On Windows, there's an arbitrary limit of 256Mb due to the way memory
is allocated, which is the primary cause of running out of heap.  On
Unix, the heap size is unlimited by default.

Fix test in findPackages

Fix --mk-dll

omit the initial foralls for :type, unless -fglasgow-exts is on

ghc -e 'System.Exit.exitWith (System.Exit.ExitFailure 99)'

should just exit 99, rather than printing out the exception.  Also
while I'm here, make all other exceptions cause 'ghc -e' to exit(1),
which is probably what we want.

remove race conditions in the recent in_haskell sanity check code

Fix endMkDependHS

Update documentation of ghc-pkg command

Change in semantics:
  - commands which only inspect the databse (list,describe,field)
    now take into account the user database unless --global is
    given.  This behaviour matches GHC, which also uses the user
    database by default.
  - However, commands which modify the database still use the
    global database, unless --user is given.

Also, allow P-* to be given as a package identifier, which means
"all versions of package P".

Be clearer about HC files: you need a set generated on the same
platform, not just similar hardware.

Improve the module-not-found error message when we didn't actually
search for it anywhere.  This happens when we know all the places it
might be: either in the current program, or a package.

cmSetContext: check whether the modules specified actually exist and
emit the usual Finder error message if any don't.

This regressed duing the ModuleName->Module changeover, where I was a
bit heavyhanded in removing moduleNameToModule: it also checked for
the existence of each module and emitted a sensible exception.

More libghccompat nonsense

-d*-lint shouldn't turn off recompilation checking

Put the GlobalRdrEnv back into the ModIface, so it gets preserved when
we reload a module without recompiling it.

Fix documentation for ghc-pkg list.

Fix help text for ghc-pkg list to match the implementation (it was
easier to do this than fix the implementation, and in fact this is
more consistent).

Catch up with mutable list changes

Careful with mutable list entries that point to THUNKs: the thunk
might be updated, and the resulting IND_OLDGEN will be on the mutable
list twice.

We previously avoided this problem by having an extra MUT_CONS object
on the mutable list pointing to the THUNK, so that we could tell the
difference between the entry on the mutable list that used to be the
THUNK, and the new entry for the IND_OLDGEN.

We don't have MUT_CONS any more (this was part of the cleanup from
separating the mutable list from the heap).  So, now, when scavenging
an IND_OLDGEN on the mutable list, we check whether it is pointing to
an already-evacuated object.  This is a bit crude, but at least it is
a localised hack.

Fix a bug: thunk_selector_depth was being incremented before checking
that we had reached the depth limit, and not decremented if we had
reached the limit.

Allow package identifiers with versions in the old command-line syntax

Add a missing delcaration for stg_MUT_ARR_PTRS_FROZEN0_entry (unbreak
the unregisterised build).

Print the banner before the "reading package config file" messages in
verbose mode.

Make sure we don't register a package with an identifier that cannot
be parsed later.

NO_INSTALL_LIBRARY=YES

Fix library installation: we now install libraries *unless*
NO_INSTALL_LIBRARY=YES.

GC changes: instead of threading old-generation mutable lists
through objects in the heap, keep it in a separate flat array.

This has some advantages:

  - the IND_OLDGEN object is now only 2 words, so the minimum
    size of a THUNK is now 2 words instead of 3.  This saves
    some amount of allocation (about 2% on average according to
    my measurements), and is more friendly to the cache by
    squashing objects together more.

  - keeping the mutable list separate from the IND object
    will be necessary for our multiprocessor implementation.

  - removing the mut_link field makes the layout of some objects
    more uniform, leading to less complexity and special cases.

  - I also unified the two mutable lists (mut_once_list and mut_list)
    into a single mutable list, which lead to more simplifications
    in the GC.

xrefs to Cabal docs

Fix previous commit. Got distracted, thought I had already tested.

Fix Mac OS X-specific breakage introduced by rev. 1.67 of ghc/includes/Stg.h.
Adding the used attribute to static functions caused gcc to emit
Apple-specific .no_dead_strip directives, which messed things up.

Add a special case for Mac OS X to FP_PROG_AR_NEEDS_RANLIB.
On Mac OS X, we need to run ranlib after installing .a files because
they contain a timestamp which will be outdated after installing.

We could do a proper test for that, but it would contain the command
'sleep 6' and still apply only to one platform.

PowerPC: mangler trouble again

PowerPC: #ifdef darwin_REGS -> #if darwin_REGS

Keep .type/.size directives for _entry symbols, but rename them to
_info.  This helps cachegrind to find where my instructions are going.

Add comment

oops, fix building of GHCi libraries.  We only want to avoid the
*install* if this isn't a package.