Re: multifd/tcp/zlib intermittent abort

[Thomas Huth]

On 15/03/2022 16.40, Daniel P. Berrangé wrote:
> On Tue, Mar 15, 2022 at 03:30:27PM +0000, Peter Maydell wrote:
> > On Tue, 15 Mar 2022 at 15:03, Peter Maydell <peter.maydell@linaro.org> wrote:
> > > Maybe we're running into this bug
> > > https://bugs.launchpad.net/ubuntu/+source/zlib/+bug/1961427
> > > ("zlib: compressBound() returns an incorrect result on z15") ?
> >
> > Full repro info, since it's a bit hidden in this long thread:
> >
> > Build an i386 guest QEMU; I used this configure command:
> >
> > '../../configure' '--target-list=i386-softmmu' '--enable-debug'
> > '--with-pkgversion=pm215' '--disable-docs'
> >
> > Then run the multifd/tcp/zlib test in a tight loop:
> >
> > X=1; while QTEST_QEMU_BINARY=./build/i386/i386-softmmu/qemu-system-i386
> > ./build/i386/tests/qtest/migration-test  -tap -k -p
> > /i386/migration/multifd/tcp/zlib ; do echo $X; X=$((X+1)); done
> >
> > Without DFLTCC=0 it fails typically within 5 or so iterations;
> > the longest I've ever seen it go is about 32.
>
> So if this is a host OS package bug we punt to OS vendor to fix,
> and just apply workaround in our CI ?  eg
>
> $ git diff
> diff --git a/.travis.yml b/.travis.yml
> index c3c8048842..6da4c9f640 100644
> --- a/.travis.yml
> +++ b/.travis.yml
> @@ -218,6 +218,7 @@ jobs:
>           - TEST_CMD="make check check-tcg V=1"
>           - CONFIG="--disable-containers 
> --target-list=${MAIN_SOFTMMU_TARGETS},s390x-linux-user"
>           - UNRELIABLE=true
> +        - DFLTCC=0
>         script:
>           - BUILD_RC=0 && make -j${JOBS} || BUILD_RC=$?
>           - |

Sounds like a good idea - but you should certainly add a proper comment 
here, too, so that we can later remind ourselves to remove the workaround again.

 Thomas