This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: ABI / dynamic list now on wiki
- From: Andreas Krebbel <krebbel at linux dot vnet dot ibm dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>, libc-alpha at sourceware dot org
- Cc: "Carlos O'Donell" <carlos at redhat dot com>
- Date: Fri, 24 Jan 2014 09:26:58 +0100
- Subject: Re: ABI / dynamic list now on wiki
- Authentication-results: sourceware.org; auth=none
- References: <Pine dot LNX dot 4 dot 64 dot 1401230114430 dot 16981 at digraph dot polyomino dot org dot uk>
On 23/01/14 02:34, Joseph S. Myers wrote:
> https://sourceware.org/glibc/wiki/ABIList
>
> now lists what I think are the 64 incompatible (architecture, ABI)
> combinations supported by glibc (soon to be 63 once we remove am33, but
> will go up as new cases such as AArch64 ILP32 (2 ABIs), and MIPS o32
> -mfp64 (4 ABIs), are added to glibc, as well as with new architectures).
>
> I've chosen to list for every ABI whether it's hard or soft float, and
> what endianness, but only to mention whether it's 32-bit or 64-bit in
> cases of 64-bit architectures (but to do so whether or not both 32-bit and
> 64-bit ABIs are supported ... in the IA64 case, ILP32 exists but not with
> Linux, and for HPPA, 64-bit exists but not for Linux userspace).
>
> Would architecture maintainers please check the lists of ABIs for their
> architectures? Especially for alpha, hppa, classic m68k, s390 and sparc
> (confirmation about non-use of soft-float configurations wanted) and for
> microblaze (confirmation about hard- and soft-float ABIs being the same
> and both endiannesses being in use wanted), but it's a good idea for other
> architectures as well.
On S/390 the soft float ABI is only used for the Linux kernel. Or perhaps in user space code to
prevent the backend from using FPRs in code not using FP types otherwise but it is no ABI change in
that case.
-Andreas-
>
> Could whoever maintains the list of Debian multiarch triplets
> <https://wiki.debian.org/Multiarch/Tuples> please make sure they are
> unambiguous regarding which ABI is referred to (when it's an ABI on this
> list)? (In particular, I note that the big-endian ARM entries there fail
> to distinguish between BE8 and BE32, but those two require completely
> separate sets of executables and shared libraries; only unlinked .o files
> are compatible between the two big-endian variants.) I'd also encourage
> the maintainer of the Linaro list to link to the glibc wiki page.
>