This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: [PATCH] Add ia64-linux and alphaev6-linux baseline_symbols.txt
- From: Phil Edwards <phil at jaj dot com>
- To: Jakub Jelinek <jakub at redhat dot com>
- Cc: bkoz at redhat dot com, libstdc++ at gcc dot gnu dot org
- Date: Thu, 5 Sep 2002 11:12:57 -0400
- Subject: Re: [PATCH] Add ia64-linux and alphaev6-linux baseline_symbols.txt
- References: <20020905104942.R1013@sunsite.ms.mff.cuni.cz>
> Two notes about ABI check:
> a) it calls sort without explicitely saying what locale it wants to be
> sorted in, i686-pc-linux-gnu files have been apparently sorted with
> LC_ALL=en_US or other locale with identical collation.
> I've followed this, but maybe it would be better to use C locale,
> ie. s/sort/LC_ALL=C sort/ and resort the files.
Might be useful, yes. (Do we really need to sort the lists at all, given
that we're storing them in maps?)
> b) maybe make abi-check should be more liberate about platform tripples,
> e.g. i386-pc-linux-gnu ought to be always identical to
> i686-pc-linux-gnu which should be identical i386-redhat-linux or
> i586-suse-linux or whatever. Similarly it shouldn't matter if the
> build is alphaev67 or alphaev5 or alpha, etc.
> But it matters if the build is sparc64- vs. sparc-...
Absolutely; this was brought up before, and some renaming is in order.
(I plan to do this all at once, and soonish.)
> Creating symlinks for every thinkable variant is lame, so I guess
> we need some config.base script or something like that that will
> take a tripplet and canonicalize it.
We already do, see configure.target. If the canonicalizationized[*] name
isn't what we want, we can override it in the last case statement in the file,
since the baseline symbols depend both on cpu and os.
Phil
[*] sorry, got started typing and didn't know when to stop
--
I would therefore like to posit that computing's central challenge, viz. "How
not to make a mess of it," has /not/ been met.
- Edsger Dijkstra, 1930-2002