[PATCH] ci: support Travis CI's aarch64 builds

[Dmitry V. Levin]

On Thu, Nov 07, 2019 at 10:41:06AM +0100, Paul Chaignon wrote:
> On Wed, Nov 06, 2019 at 02:07:47AM +0300, Dmitry V. Levin wrote:
> > On Tue, Nov 05, 2019 at 07:43:07PM +0100, Paul Chaignon wrote:
> > > Travis CI now supports aarch64 builds, launched with the new "arch: arm64"
> > > tag.  These builds are running inside LXD only for the moment.
> 
> [...]
> 
> > > diff --git a/ci/run-build-and-tests.sh b/ci/run-build-and-tests.sh
> > > index f52bf5cd..1f1de42d 100755
> > > --- a/ci/run-build-and-tests.sh
> > > +++ b/ci/run-build-and-tests.sh
> > > @@ -27,6 +27,12 @@ case "${TARGET-}" in
> > >  		CC="$CC -m32"
> > >  		DISTCHECK_CONFIGURE_FLAGS="$DISTCHECK_CONFIGURE_FLAGS --build=i686-pc-linux-gnu --target=i686-pc-linux-gnu"
> > >  		;;
> > > +	aarch64)
> > > +		# The new Travis CI aarch64 builds under LXD have rubbish data in
> > > +		# the /proc/[pid]/stat file required by strace-DDD.test.
> > > +		sed -ri 's/(XFAIL_TESTS_ =.*)$/\1 strace-DDD.test/' tests/Makefile.am
> > > +		DISTCHECK_CONFIGURE_FLAGS="--disable-mpers"
> > 
> > I think we should use --enable-mpers=check instead,
> > otherwise all this dance around gcc-multilib-arm-linux-gnueabihf
> > has little meaning.
> 
> I've made the change, but I'm not sure I understand the relation between
> --enable-pers and the packages installed for the CI...?

gcc-multilib-* packages is the Debian way of installing gcc multilib
support required for --enable-mpers.

> > Btw, there is an unfinished work in esyr/aarch64 branch to have
> > a usable mpers on aarch64.
> 
> I'll have a look.
>  
> > Wrt strace-DDD.test, do you think we could diagnose the oddness
> > inside the test and framework_skip_ it in a usual way?
> 
> It looks like a bug on Travis CI's side (I can't reproduce locally under
> LXD).  I've reached out to them (through the community forum for now [1]);
> maybe we should wait to hear from them before encoding this behavior into
> the test?  I can send a second patch to "fix" the test once we know more,
> or we can put the present patch on hold until then.
> 
> 1 - https://travis-ci.community/t/unexpected-zero-values-in-proc-pid-stat/5774

Apparently, these zero process group ID and session ID could be allowed in
strace-DDD.test as they don't interfere with the purpose of that test.

I've pushed a commit to address this issue.

Also, I've pushed your commit that makes umount2.test work properly
when umount2 syscall is filtered using seccomp-bpf.


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20191108/2eeadfb8/attachment.bin>