[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gforth] ABI-CODE broken for git-head
From: |
David Kuehling |
Subject: |
Re: [gforth] ABI-CODE broken for git-head |
Date: |
Thu, 27 Mar 2014 06:25:45 +0100 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/23.4 (gnu/linux) |
>>>>> "Bernd" == Bernd Paysan <address@hidden> writes:
> Am Mittwoch, 26. März 2014, 03:00:44 schrieb David Kuehling:
>> Hi,
>>
>> ABI-CODE seems to be broken with git HEAD, at least on amd64 and
>> mipsel.
> Yes, there was still a stub in the ABI-CODE vtcopy,, so it didn't copy
> the correct compile, action. Fixed that.
> BTW: IMHO, this abi-code-test.fs stuff should be part of the test
> cases. Check MACHINE to select the right assembler code.
MACHINE may be insufficient in cases where we have multiple possible
ABIs? Like mipsel o32 vs. n32 vs. eabi? Similar problems on ARM (and
nowadays with i386 vs. x32?). Didn't even know of MACHINE, we should
mention it in the assembler code chapter of the manual (ha, finally
inline assembler being usable in "production-code" Gforth programs!).
Just tried MACHINE TYPE on mipsel oabi32: just outputs "mips". Looked
it up, this is generated from configure via config.status substituting
stuff in envos.fs.in. It doesn't seem that config.status knows any
helpful CPU architecture details. Maybe we'll have to do it the hard
way and check for MIPS_* #defines during configure (or compile-time) and
somehow export this information (depending on architecture) to Gforth.
David
--
GnuPG public key: http://dvdkhlng.users.sourceforge.net/dk2.gpg
Fingerprint: B63B 6AF2 4EEB F033 46F7 7F1D 935E 6F08 E457 205F
pgprwBYBuzyvh.pgp
Description: PGP signature