GitLab

Haskell quiz:  what should this program do?

  main = return undefined

answer: run to completion and exit successfully.  GHC erroneously
evaluates the returned value from main, which causes this example to
fail with an uncaught exception (the evaluation is even done outside
of the main exception handler!).

Support for contexts on data types and records from Derek Elkins.

Fix tryM for GHC 5.00.x and 5.02.x.

Add comment about lack of instance-support in .hi-boot

- Remove mention of MutableByteArray and ByteArray

- Add section on memory allocation in the FFI, with emphasis on how
  GHC's implementations of the various kinds of allocation perform.

----------------------------------
      Fix a long-standing egregious parser bug
	----------------------------------

	*** MERGE TO STABLE ***

	*** NB: the important part of this commit
	***	got committed by accident with an
	***	unrelated message.  This commit
	***	should be from rev 1.50 to 1.52 of RdrHsSyn


GHC has parsed

	data T String = T String

without complaint, ever since day 1!  This led to consequential
incomprehensible messages.  The fix is easy.

Document external core a bit more

Restructure documentation about undecidable instances

Buglet in external-core parsing

Use STK_CHK_NP rather than STK_CHK_GEN in the entry code for AP and
AP_STACK thunks.

The reason is that if the check fails, on return the closure might
have been blackholed (perhaps due to lazy blackholing, or even because
another thread has entered it), so we *must* enter the closure to
continue, rather than just continuing at the original entry point.

All compiler-generated thunks do this anyway, I think these are the
only cases of thunks which don't follow the rule.

Thanks to Robert Ennals for spotting, and correctly diagnosing, the
bug.

Fix typo

Fix compilation problem on mingw32

Minor wording

Update docs on class method types

Markup glitch

Oops, byteCodeGen should get the bindings from CorePrep, not CoreTidy
(should fix last night's GHCi breakage).

some rearrangement.

also install libraries.html

notes for Hugs

generate docs/libraries.html

Split off the hierarchical libraries page, so the Hugs folks can use it too.

Duh.  hsc2hs should be in $(GhcBinDistShScripts), not
$(GhcBinDistBins), otherwise it doesn't get the right directories
tacked on the front at installation time.  Strange that nobody
complained that hsc2hs wasn't working from a binary dist *shrug*.

Make it compile again with 5.04.x

Fix compilation breakage with GHC 4.08.x.

win32/Ticker.c:startTicker(),
win32/IOManager.c:NewIOWorkerThread():
  pass in non-NULL pointer to _beginthreadex()'s outgoing thread ID arg.
  Needed to work with Win9x's impl of CreateThread().

A round of space-leak fixing.

  - re-instate zapping of the PersistentCompilerState at various
    points during the compilation cycle in HscMain.  This affects
    one-shot compilation only, since in this mode the information
    collected in the PCS is not required after creating the final
    interface file.

  - Unravel the recursive dependency between MkIface and
    CoreTidy/CoreToStg.  Previously the CafInfo for each binding was
    calculated by CoreToStg, and fed back into the IdInfo of the Ids
    generated by CoreTidy (an earlier pass).  MkIface then took this
    IdInfo and the bindings from CoreTidy to generate the interface;
    but it couldn't do this until *after* CoreToStg, because the CafInfo
    hadn't been calculated yet.  The result was that the CoreTidy
    output lived until after CoreToStg, and at the same time as the
    CorePrep and STG syntax, which is wasted space, not to mention
    the complexity and general ugliness in HscMain.

    So now we calculate CafInfo directly in CoreTidy.  The downside is
    that we have to predict what CorePrep is going to do to the
    bindings so we can tell what will turn into a CAF later, but it's
    no worse than before (it turned out that we were doing this
    prediction before in CoreToStg anyhow).

  - The typechecker lazilly typechecks unfoldings.  It turns out that
    this is a good idea from a performance perspective, but it also
    means that it must hang on to all the information it needs to
    do the typechecking.  Previously this meant holding on to the
    whole of the typechecker's environment, which includes all sorts
    of stuff which isn't necessary to typecheck unfoldings.  By paring
    down the environment captured by the lazy unfoldings, we can
    save quite a bit of space in the phases after typechecking.

Use listArray rather than array in (un)?boxedTupleArr.

Remove unused export.

underscore some variables that aren't used.

Compilation fix after recent addition of environment stuff to System.Posix.

----------------------------------
	Improve higher-rank type inference
	----------------------------------

Yanling Wang pointed out that if we have

	f = \ (x :: forall a. a->a). x

it would be reasonable to expect that type inference would get the "right"
rank-2 type for f.  She also found that the plausible definition

	f :: (forall a. a->a) = \x -> x

acutally failed to type check.

This commit fixes up TcBinds.tcMonoBinds so that it does a better job.
The main idea is that there are three cases to consider in a function binding:

  a) 'f' has a separate type signature
	In this case, we know f's type everywhere

  b) The binding is recursive, and there is no type sig
	In this case we must give f a monotype in its RHS

  c) The binding is non-recursive, and there is no type sig
	Then we do not need to add 'f' to the envt, and can
	simply infer a type for the RHS, which may be higher
	ranked.

Remove mention of man pages, which we don't have, and clarify a couple
of other things while I'm here.

*** My last commit had a bogus commit message **
	*** This is the right one.  I've appended the **
	*** list of files changed as well.

	-------------------------------------
	   Fix up HOST/TARGET_OS ifdefs
	-------------------------------------

Sigbjorn and Simon helped me to understand when it's appropriate to
say 'ifdef mingw32_HOST_OS' and when 'ifdef mingw32_TARGET_OS'.  I've
documented this in the building guide.

I've fixed up the ifdefs I erroneously changed last week, in this
commit.  I'm not certain about all of them.  For examples, in
DriverPhases there's an ifdef that deals with filename suffices.
Should this be the filename suffices for the system on which GHC runs
(HOST), or the system on which the compiled code will run (TARGET)?
Currently it's the latter.

Similarly, there's some path mangling ('/' vs '\') in DriverState:
should this be for the system on which GHC runs (HOST) or the system
on which the compiled code will run (TARGET)?  Currently the former.

I think some of the darwin_TARGET_OS should be darwin_HOST_OS's.

Finally, for the GHCi stuff we require that HOST_OS = TARGET_OS.  Rather than
make an arbitrary choice in (e.g.) Linker.lhs, it's be better to say
'#ifdef mingw32_OS', and arrange that this macro is set only in stage-2.
I have not done this.

I think these are 'right'.  But I'm not sure the filename mangling is
consistent throughout.

[Changes in the original commit, which had a bogus message.]

  Revision  Changes    Path
  1.46      +30 -0     fptools/docs/building/building.sgml
  1.148     +4 -4      fptools/ghc/compiler/ghci/InteractiveUI.hs
  1.114     +2 -2      fptools/ghc/compiler/main/DriverFlags.hs
  1.23      +3 -1      fptools/ghc/compiler/main/DriverPhases.hs
  1.85      +11 -11    fptools/ghc/compiler/main/SysTools.lhs
  1.20      +3 -3      fptools/ghc/compiler/utils/Panic.lhs

Three Template Haskell improvements

a) Add type synonyms to THSyntax (and DsMeta, Convert)

b) Make Q into a newtype instead of a type synonym

c) Eliminate tiresome and error prone argument to DsMeta.wrapGenSyms
   and similarly addTyVarBinds

bring getourtimeofday() proto into scope

Clean up code&interfaces that deals with timers and asynchrony:

- Timer.{c,h} now defines the platform-independent interface
  to the timing services needed by the RTS. Itimer.{c,h} +
  win32/Ticker.{c,h} defines the OS-specific services that
  creates/destroys a timer.
- For win32 plats, drop the long-standing use of the 'multimedia'
  API timers and implement the ticking service ourselves. Simpler
  and more flexible.
- Select.c is now solely for platforms that use select() to handle
  non-blocking I/O & thread delays. win32/AwaitEvent.c provides
  the same API on the Win32 side.
- support threadDelay on win32 platforms via worker threads.

Not yet compiled up on non-win32 platforms; will do once checked in.

Typos

Fix buglets in new unused-binding story

Use old-style foreign import for bkwd compatibility

-------------------------------------
	Improve the "unused binding" warnings
	-------------------------------------

We've had a succession of hacks for reporting warnings for
unused bindings.  Consider

	module M( f ) where

 	f x = x

	g x = g x + h x
	h x = x

Here, g mentions itself and h, but is not itself mentioned. So
really both g and h are dead code.  We've been getting this wrong
for ages, and every hack so far has failed on some simple programs.

This commit does a much better job.  The renamer applied to a bunch
of bindings returns a NameSet.DefUses, which is a dependency-ordered
lists of def/use pairs.  It's documented in NameSet.
Given this, we can work out precisely what is not used, in a nice
tidy way.

It's less convenient in the case of type and class declarations, because
the strongly-connected-component analysis can span module boundaries.
So things are pretty much as they were for these.


As usual, there was a lot of chuffing around tidying things up.
I havn't tested it at all thoroughly yet.

Various unrelated import-decl-pruning has been done too.