On Sat, Dec 28, 2024 at 08:32:13PM +0100, Jeremie Courreges-Anglas wrote:
> On Sat, Dec 28, 2024 at 03:53:52PM +0100, Mark Kettenis wrote:
> > > Date: Fri, 27 Dec 2024 18:01:44 +0100
> > > From: Jeremie Courreges-Anglas <jca@wxcvbn.org>
> > >
> > > I'm working on an update to gdb-15.2, but the changes below are
> > > already legit.
> > >
> > > 0. add some missing files to ALLDEPFILES for consistency, update
> > > comment
> > > 1. add ppc64-obsd-tdep.o to ALL_64_TARGET_OBS, so that it is built on
> > > all architectures and not only on powerpc64. This makes it possible
> > > to read powerpc64 core files on other architectures. However...
> > > 2. hooking up this file results in a runtime assertion because of a
> > > conflict between powerpc and powerpc64 (see the patch for the error
> > > message). This error is probably already breaking gdb-13.2 runtime
> > > on powerpc64 since Makefile rev 1.90 where I actually activated
> > > --enable-targets=all.
> > > 3. drop patches/patch-gdb_doc_gdb_texinfo: it's not needed since we
> > > patch the existing .info file, and it breaks with gdb-15.2 where the
> > > build system tries to regenerate the .info
> > >
> > > gkoehler & kettenis: if you could please check the current egdb
> > > runtime on powerpc64 (point 2 above), and whether the diff below fixes
> > > it. ... are C++ exceptions still broken on that arch? :-/
> > >
> > > ok?
> >
> > Well, the current port doesn't even build on powerpc64 because of
> > undefined references. Probably this is what point 1 is fixing.
>
> Hmm, that's not what I expected; maybe fixing --enable-targets=all had
> even more consequences than I thought?
Indeed point 1 should fix the build on powerpc64 after
devel/gdb/Makefile rev 1.90. gdb/Makefile.in clearly states:
# NOTE: Every file added to a gdb_target_obs variable for any target here
# must also be added to either ALL_TARGET_OBS or ALL_64_TARGET_OBS
# in Makefile.in!
With --enable-targets=all, those lists are the canonical source of
*tdep.o files to build, clobbering what configure.tgt detects, so they
indeed must be accurate.
I'm going to commit this soonish.
--
jca
No comments:
Post a Comment