Re: Bug#873465: apt: wrongly reports failure on 'update' on mips64el, triggers d-i FTBFS
On Mon, Aug 28, 2017 at 07:50:34AM +0200, Cyril Brulebois wrote:
> Control: tag -1 patch
>
> Cyril Brulebois <kibi@debian.org> (2017-08-28):
> > Given apt 1.5~beta2 was uploaded on the evening of the 17th, it probably
> > reached the archive on the 18th, and was first used in the build chroot
> > for the 0030Z build on the 19th. I'm filing this against apt since it
> > seems clear to me there were no actual download errors, but it also
> > seems likely that apt might be encountering gcc-7 related
> > miscompilations, see #871514.
> >
> > In any case, a speedy fix would be much appreciated, as this is a
> > blocker for the D-I Buster Alpha 1 release (which has already been
> > delayed way too much for other reasons).
>
> I've built apt with -O0 (it seems a more acceptable workaround than
> going back to gcc-6), and setting PATH/LD_LIBRARY_PATH to point to its
> debian/tmp directory, I've managed to get rid of the issue I was getting
> when trying to build debian-installer, and even managed to get a full
> build.
>
> Of course this generates a lot of noise in dpkg-gensymbols (end of the
> build log attached), but since it didn't bail on me, I guess that's OK?
>
> Trivial patch for debian/rules attached; of course this can be limited
> to mips64el, but I'd like to get a maintainer's opinion as to whether
> that's an acceptable workaround for the time being. If it is, I can NMU.
Hmm, shouldn't it also STRIP the -O2? I guess this would work too and
might be nicer?
ifeq ($(DEB_HOST_ARCH),mips64el)
DEB_BUILD_OPTIONS += noopt
endif
(noopt also changes -O2 into -O0). Anyhow, feel free to upload an
NMU with ifeq ($(DEB_HOST_ARCH),mips64el) around either change.
--
Debian Developer - deb.li/jak | jak-linux.org - free software dev
| Ubuntu Core Developer |
When replying, only quote what is necessary, and write each reply
directly below the part(s) it pertains to ('inline'). Thank you.
Reply to: