This is the mail archive of the
mailing list for the glibc project.
Re: [PATCH] PowerPC64 ELFv2 ABI 2/6: Remove function descriptors
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: joseph at codesourcery dot com (Joseph S. Myers)
- Cc: amodra at gmail dot com, libc-alpha at sourceware dot org
- Date: Wed, 13 Nov 2013 19:17:08 +0100 (CET)
- Subject: Re: [PATCH] PowerPC64 ELFv2 ABI 2/6: Remove function descriptors
- Authentication-results: sourceware.org; auth=none
Joseph Myers wrote:
> On Wed, 13 Nov 2013, Ulrich Weigand wrote:
> > However, there are assembler files where none of this applies. This is
> > typically files that define only data objects. We had thought to maybe
> > force these to use .abiversion, but it turns out that this would have
> > required annoying changes (even in glibc we have platform-independent
> > assembler source files defining data --- these would have had to be
> > patched with powerpc-specific code), for no real good reason: such
> > files (data only) are actually compatible with either ABI anyway!
> On ARM, we found that putting .eabi_attribute directives in sysdep.h (to
> specify required and maintained stack alignment, a similar case where
> default assembler output isn't compatible with compiler output) covered
> almost all assembly sources in glibc, leaving only a handful of .S files
> needing such directives directly in the .S file.
I guess it couldn't hurt to put a .abiversion in sysdep.h, thanks for
the tip! But the problem isn't really glibc in itself, I was just
using this as example.
In the end we'll have to rebuild a whole distribution for powerpc64le,
and we'd really like to avoid anything that makes this harder than it
already is. Having a mandatory requirement that all assembler files
*must* be marked doesn't seem a good idea to me (in particular if there
is no real good reason anyway, as mentioned above).
Dr. Ulrich Weigand
GNU/Linux compilers and toolchain