2016-09-04 05:35 Aaron M. Ucko:
Aptitude recently started aborting under some circumstances, notably
when asked to calculate safe upgrades on my multiarch-enabled system.
The aborts turned out to occur on line 1182 of src/generic/apt/apt.cc,
with t1 values that didn't match any cases because the ArchSpecific
flag was originally set and never masked out. I've gotten good
results by masking out *all* high bits when setting t1 and t2, per the
below patch, but haven't examined the code closely enough to be
certain that it's correct or sufficient.
Could you please take a look?
Thaks for the analysis and the patch. I'm a bit busy but will try to
take a deeper look soonish.
On a first glance, using the module operator rather than bitwise
operations looks a bit odd.
Also, the code is from before 2005 (the beginning of the VCS history) so
it's a bit strange that it survived all these years if it produces
crashes, but then again maybe the wrong behaviour it's caused by more
recent changes to support multiarch, or changes in apt, etc.
Manuel A. Fernandez Montecelo <manuel.montez...@gmail.com>
Aptitude-devel mailing list