Re: libreoffice-5-0 failure in [ CUT ] libreofficekit_tiledrendering

On Thu, Jun 18, 2015 at 04:23:36PM +0200, Lionel Elie Mamane wrote:
> On Mon, Jun 15, 2015 at 02:32:23PM +0200, Miklos Vajna wrote:
> > On Wed, Jun 03, 2015 at 02:54:45PM +0200, Lionel Elie Mamane <lionel@mamane.lu> wrote:
> > > > What are those autogen.input options? Can you follow the instructions of
> > > > the failed test and get a backtrace?
> > >
> > > Both are attached. To me the backtrace looks unrelated to "tile
> > > rendering", but then this is the only failing thing in a
> > >  make check -k
> >
> > I only have an implicit --without-system-libs dbgutil build from
> > libreoffice-5-0, and there I don't see this problem. It would be
> > interesting to know if the failure is indeed conditional on
> > --with-system-libs or it's something else.
> >
> > An other approach is to see if the branch point was good for you, if so,
> > then bisect what commit broke the test with your build options.
>
> Nope, branch point is not good. Given that a recompile takes hours,
> I'm rather unlikely to pursue this in detail. I started a "git
> bisect", let's see if I have enough patience to get to the end.

Grrr... "git bisect" refuses to work in the direction "find the commit
that fixed this", it insists to do "find the commit that broke
this". So I cannot bisect between (known bad) branch point and (known
good) master.

-- 
Lionel