On 07/30/2012 12:18 AM, Philipp Kern wrote: > On Sun, Jul 29, 2012 at 10:16:57PM -0700, tony mancill wrote: >> nmu bobcat_3.01.00-1 . ALL . -m "recompilation with current g++ (closes: >> #683049)" > > What's the condition here? What changes to g++ will cause it to need a > rebuild? > > Kind regards > Philipp Kern
Fair question. At this point I only have empirical evidence that the
segfault problems with bobcat on amd64 disappear after a rebuild, and I
lack sufficient expertise to ferret out why or exactly when.
The original bug reported is a FTBFS [1] for a package that uses
libbobcat-dev, which I cloned ([2]). Those bugs contain information
about the original problem, and I'm adding more here after an exchange
with the upstream author. The source code in bobcat that is failing
with the binary package in testing/sid is below
(bobcat/pattern/position.cc):
Pattern::Position Pattern::position(size_t index) const
{
regoff_t begin;
return
(
index >= d_beyondLast ||
(begin = d_subExpression[index].rm_so) == -1
) ?
Position(string::npos, string::npos)
:
Position(begin, d_subExpression[index].rm_eo);
}
Rebuilding bobcat in a sid chroot resolves the issue. So it could have
been a build env problem when the bobcat binaries were built and
uploaded, and it's also possible that the problem is isolated to amd64.
In either case, it seemed best to request that the package be rebuilt
on the Debian buildds.
Thank you,
tony
[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=682640
[2] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=683049
signature.asc
Description: OpenPGP digital signature
