On Mon, Jan 5, 2009 at 7:10 PM, David Chisnall <[email protected]> wrote:
> Hi everyone,
>
> Truls reports that NSRanges are now being correctly boxed on Linux.
> My ABI-conformance code is very simple and won't support some of the
> weird and wonderful ABIs out there (for example, HP-UX has different
> conventions for complex numbers and structures that are equivalent to
> them), and might not work for some large structure types.  It should,
> however, now work for the NS* structs on most common platforms.
>
> Whether a structure is returned on the stack or in registers is
> defined by some numbers in ABI.h.  I'm not sure what these should be
> for x86-64 (although the good news is that, unlike x86-32, everyone
> uses the same calling conventions there).
>
> I think this is the last major issue I wanted to fix before 1.0.  Non-
> local returns now work for the static compiler, but not for the JIT.
> I am not sure if this is my bug which is only appearing in the JIT, or
> an LLVM bug, but I'm happy to leave fixing it for 2.0 (non-local
> returns are usually bad programming style anyway).
>
> Please can everyone test this before I move it into stable.

I did a quick test. nsrange.st in Compiler/examples/ does work now,
but mult.st failed (segment fault).
So I run the script in SmallkitTests/ and three tests failed:

------------------------------------------------------
Test: TestPrototypes
Aborted
-e TestPrototypes: FAIL (crash)
------------------------------------------------------
Test: TestJustAnInteger
Segmentation fault
-e TestJustAnInteger: FAIL (crash)
------------------------------------------------------
Test: TestBlockReturn
Segmentation fault
-e TestBlockReturn: FAIL (crash)

llvm is updated last December after 2.4 release and Etoile is the latest.

Yen-Ju

>
> David
>
>
> P.S. In tangentially related news, the paper about the Étoilé runtime
> was just published:
>
> http://www.jot.fm/issues/issue_2009_01/article4/
>
> Thanks to everyone who helped with that,
>
>
> _______________________________________________
> Etoile-discuss mailing list
> [email protected]
> https://mail.gna.org/listinfo/etoile-discuss
>
_______________________________________________
Etoile-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/etoile-discuss

Répondre à