Re: Accepted gcc-defaults 1.118 (source all amd64)
Hi,
> GCC 4.7 was blamed for this, but now I'm looking back and wondering if
> that was fair, and if it was such a bad thing all of this happened.
>
> The solution in gpsd VCS seems to have been applying a rather large
> changeset, fixing things from Coverity static analysis, potential buffer
> overruns and simple code bugs:
>
> http://git.recluse.de/?p=debian/pkg-gpsd.git;a=commitdiff;h=694d9866388d3c99ea5bdf2d570865f0bfa79d44
no, that was not the solution. A lot of those fixes were more or less cosmetic
or fixes for minor bugs.
> It seems the change of compiler tickled one of these bugs, made it
> noticable (causing a segfault, instead of some other behaviour, which
> was maybe correct or perhaps not), and then it got fixed.
The compiler indeed made one uninitialized integer show up, leading to an array
access at rather random positions. Its only a few lines diff.
http://git.recluse.de/?p=debian/pkg-gpsd.git;a=blob;f=debian/patches/d5fc020_Explicitly-initialize-AIS-queue-pointer;h=07b437f03dfff25bd228d830471633f09c42795c;hb=694d9866388d3c99ea5bdf2d570865f0bfa79d44;js=1
> So I think at least some good may have come out of the change of
> compiler, even if it is decided to roll back gcc-defaults. Most of the
> work to at least get packages ready for GCC-4.7 seems done now, and
> maybe the same fixes will help if there's some future switch to LLVM/Clang.
Yeah, I'm not saying that finding this issue was bad - but I think it could have
been done after the release of wheezy. Especially in this case it made a lot of
work to actually find the issue - and the normal user would probably never run
into it. I would have preferred to spend my time on more important fixes for Wheezy.
Cheers,
Bernd
--
Bernd Zeimetz Debian GNU/Linux Developer
http://bzed.de http://www.debian.org
GPG Fingerprint: ECA1 E3F2 8E11 2432 D485 DD95 EB36 171A 6FF9 435F
Reply to: