On Thu, 2011-08-25 at 13:38 -0400, Stephen Gallagher wrote:
On Thu, 2011-08-18 at 17:01 +0200, Jan Zelený wrote:
> > Dne 17.8.2011 15:29, Jan Zelený napsal(a):
> > >>> Patch #0006:
> > >>> -<nitpick> please just for safety use parentheses in
ternary
> > >>> operator (line 92 and couple more similar)</nitpick>
> > >>
> > >> Done and also moved to macro.
> > >
> > > You still don't have parentheses around the condition in ternary
> > > operator. No big deal though, it doesn't concern me in the macro
since
> > > bringing an error to one possible place is less likely than to one of
> > > many possible places.
> >
> > Sorry about that. Now it's done :-)
>
> Oh, I just noticed what originally happened: you put the parentheses around
> the entire ternary operator instead of the condition as I had in mind. However
> I have no problem with the way it's done now.
>
> >
> > > One more thing: I don't like the invalidate macro. The assignment
with
> > > the comment you provided is clear enough I think.
> >
> > Removed.
> >
> > I also noticed that printf provides # flag, that adds 0x when you want
> > output in hex. So I changed 0x%.4x to %#.4x.
>
> I told myself that there must be something like that ;-) Thanks for saving me
> the work and finding that out on your own.
>
>
> Ack to all 6 patches
Nack. The debug tests cause a buffer overflow when run during rpmbuild
(I don't know why it's peculiar to rpmbuild) See attached log.
Also, please squash the attached patch into the debug tests. I tried
using valgrind to identify the issue above and found several memory
issues, which that patch corrects. However, it still doesn't solve the
rpmbuild issue.
Thanks to help from Alexander and Simo, I was able to track down the
issue. Please review the revised patch and I'll push all of these.
This patch results in a clean valgrind run of ./debug-tests as well as
resolving the buffer-overrun crash during rpmbuild.