This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: stackable abilist files
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Mike Frysinger <vapier at gentoo dot org>
- Cc: libc-alpha at sourceware dot org
- Date: Sat, 21 Apr 2012 09:07:36 +0000 (UTC)
- Subject: Re: stackable abilist files
- References: <201204210032.55249.vapier@gentoo.org>
On Sat, 21 Apr 2012, Mike Frysinger wrote:
> although, for some of these (like libc.abilist), it's hard to say that forcing
> an arch to copy the entire file is a good thing considering 99% of the over
> 2000 lines in there are going to be duplicate. maybe it'd be better to let
> these particular files hold all target tuples whether they be in ports/ or the
> main tree ?
I think the right thing is to put files (which would no longer contain
regular expressions for architecture triplets at all) in sysdeps
directories (e.g. sysdeps/unix/sysv/linux/i386/abilist) and just take the
files from the first such sysdeps directory found. There would be
duplication, as there is with libm-test-ulps, although in most cases it
would be possible for the person adding a new API to do the mechanical
updates for all targets; I don't think that duplication is a major
problem.
I think check-abi should be part of the default "make check", and should
fail if no baselines are found in sysdeps (the expectation being that each
architecture's maintainers should add baselines if not already present ...
it appears the present baselines are missing sparc).
--
Joseph S. Myers
joseph@codesourcery.com