Re: [Valgrind-users] Valgrind-3.15.0.RC1 is available for testing

On Tue, 2019-04-09 at 09:02 +0200, Mark Wielaard wrote:
> Hi,
> 
> On Mon, 2019-04-08 at 11:11 +0200, Julian Seward wrote:
> > A first release candidate for 3.15.0 is available at
> > https://sourceware.org/pub/valgrind/valgrind-3.15.0.RC1.tar.bz2
> > (md5 = 56d9f5e25615d48110da0aa5764d481e)
> > 
> > Please give it a try on platforms that are important for you.  If
> > no serious
> > issues are reported, the 3.15.0 final release will happen on 12
> > April, that
> > is, this coming Friday.
> 
> For users are binary rpms for fedora and centos in koji and on copr
> if
> people find that easier to use for testing:
> 
> https://koji.fedoraproject.org/koji/buildinfo?buildID=1246875
> https://copr.fedorainfracloud.org/coprs/mjw/valgrind-3.15.0/
> 
> For developers the build.logs do contain the make regtest diffs for
> any
> failing tests.

I did new builds with the one extra commit and a couple of bug fixes:

https://koji.fedoraproject.org/koji/buildinfo?buildID=1247329
https://copr.fedorainfracloud.org/coprs/mjw/valgrind-3.15.0/build/881204/

This includes the following extra fixes:

valgrind commit b2d2da64b0de1c4d657b63187967b68606e84711
GET_STARTREGS for s390: fix register constraint

KDE#406352
RC1 fails cachegrind/callgrind ann tests because of missing a.c

KDE#406360
memcheck/tests/libstdc++.supp needs more supression variants

KDE#406354
dhat is broken on x86 (32bit)

KDE#406355
mcsignopass and mcsigpass fails due to a difference in gdb output

KDE#406357
RC1 fails gdbserver_tests because of gdb output change

I hand tested them all (because I made the mistake of only running the
nonexp-regtest (which doesn't include gdbserver_tests) and messed up
the data on the a.c file, which generates a warning, but shows the fix
itself is good.

May I push these fixes to git?
Should they go on the master branch, or is there a 3.15.0 branch
already?

Thanks,

Mark