GitLab

Comments

Document infix type operators

Add parser support for infix type-variable operators

Fix the superclass translation for instance decls
			Merge to STABLE

There is a long-standing difficulty whereby it's surprisingly easy 
to accidentally generate an entirely-bogus recursive dictionary when 
generating the definitions for the superclasses of an instance decl.

The problem arises because the default story is that whenever we
add a constraint to our pile of solved constraints, we automatically
add all its superclasses.  But that is simply wrong when we are trying
to generate superclasses. 

Solution: do no auto-superclass addition when solving the superclass
constraints of an instance declaration.  I think should fix it once and
for all.  

	tcrun021, tcrun033 are test cases

tcrun033 showed up the bug; thanks to Simon Foster and Ralf Laemmel.

Add missing provenance for mfix; pls merge

Add notes about implicit parameters; pls merge

Use a different magic number (0x1face64) for 64-bit interface files.
This will prevent us trying to read the dictionary out of a 32-bit
interface file on a 64-bit machine.

revert previous change, it didn't work

Retain all CAFs when dynamic Haskell libraries are used from GHCi.
The Linker usually replaces references to newCAF with references to newDynCAF,
but the system dynamic linker won't do that for us.

Also, the situation is slightly different - we never want CAFs from dylibs
to be reverted, because the dylibs might be used both by the interpreted
program and by GHCi itself.

So instead of just caf_list, there's now both caf_list and revertible_caf_list.
newDynCAF adds a CAF to revertible_caf_list, and newCAF either adds the CAF
to caf_list or to the mutable list, depending on whether we are in GHCi.

This hack is only active when Linker.c has loaded libHSbase_dyn.[so|dylib],
but for now, it applies to all CAFs, not just dynamically-linked ones.
If that is worth fixing, we could do that by checking whether the the CAF
closure or it's info pointer is in the main executable's address range.

MERGE TO STABLE

Some minimalistic documentation for -fPIC and -dynamic.

MERGE TO STABLE

give 'runhaskell' the 'exeext' treatment too

explicitly use 'exeext' to avoid cygwin 'cp' weirdity

decode OccNames in dumpStyle too

Fix something that's been bugging me for a while: by default, -ddump-*
output doesn't include uniques when it outputs internal names, but in
most cases you need them because the output hasn't been tidied, so you
end up doing -dppr-debug which is overkill.

Now, -ddump-* prints uniques for internal names by default.  This
shouldn't affect anything else.

Install runghc too

Fix what looks like a typo in the previous commit

HACK HACK HACK on x86_64

we need 16-byte aligned constants on this platform sometimes.  Don't
just 16-byte align everything, but try to detect 16-byte constants and
align just those.

All the codegen tests now go through on registerised x86_64, I'm
building a stage2 GHC now.

Add runhaskell as an optional binary

Build & install runhaskell, but only install it if there isn't already
a $(bindir)/runhaskell.

x86_64: the assembler doesn't like generating 8-byte relative
relocations between text and rodata symbols.  Hack around this by
putting SRTs in the text segment for now.

Avoid losing location info for ghci; please merge

Print full instances in ghci; merge

Comments

Better printing of types; merge please

Update comment

x86_64: Add __DISCARD__() function call to the tailcall sequence to
work around bugs in gcc (see comment for details).

Add notes about newtype deriving

Mac OS X:
Kill HaskellSupport.framework.
Instead, look for GMP.framework (a framework-version of libgmp), else look
for a normal -lgmp as usual.
The other part of HaskellSupport.framework, dlcompat, is no longer needed
(as of Mac OS X 10.3, it's included in libSystem).
It's enough to just use the normal configure tests for -ldl.

MERGE TO STABLE

Mach-O Linker: eradicate some warnings

MERGE TO STABLE

-----------------------------------------
       Make sure that LiberateCase only binds Internal Names
	-----------------------------------------

	Merge to STABLE

The nested bindings generated by LiberateCase should be Internal as
well as not-exported.  Otherwise an External Name can float to top level,
where it might name-clash with another one.

-----------------------------------------
       Fix a long-standing indirection-zapping bug
	-----------------------------------------

	Merge to STABLE

Up to now we zap indirections as part of the occurence analyser.
But this is bogus.  The indirection zapper does the following:

	x_local = <expression>
	...bindings...
	x_exported = x_local

where x_exported is exported, and x_local is not, then we
replace it with this:

	x_exported = <expression>
	x_local = x_exported
	...bindings...

But this is plain wrong if x_exported has a RULE that mentions
something (f, say) in ...bindings.., because 'f' will then die.

After hacking a few solutions, I've eventually simply made the indirection
zapping into a separate pass (which is cleaner anyway), which wraps the
entire program back into a single Rec if the bad thing can happen.

On the way I've made indirection-zapping work in Recs too, which wasn't the
case before.

* Move the zapper from OccurAnal into SimplCore
* Tidy up the printing of pragmas (PprCore and friends)
* Add a new function Rules.addRules
* Merge rules in the indirection zapper (previously one set was discarded)

Include WORD_SIZE_IN_BITS in the interface header, and test it when
reading.  Fixes a problem whereby GHC on a 64-bit platform will crash
if it tries to read an interface file generated by the same version of
GHC on a 32-bit platform.

-----------------------------------------
       Fix scoping bug for quantified type variables
	-----------------------------------------

	Merge to STABLE

When instantiating a declaration type signature, make sure to instantiate
fresh names for non-scoped type variables, else they may be spuriously shared.
Turns out that the test lib/Generics/reify tests this, which is good.

Comments are with TcMType.tcInstSigType

Try once more to get this #include right, and add a comment

ghc_boot_platform.h is the right way to get hold of the platform
defines inside the compiler.

Mention not to use comments on the same line as OPTIONS_GHC

Fix validate bug

Add missing hs_* symbols

Nuked dead code. Now the RTS is free of warnings, except for failed
inlinings of thread_obj, processHeapClosureForDead, pop, and
push. Should we remove their "inline" modifier?

Clean up Version.hs for distclean target