On Mar 22, 2014, at 8:41 AM, Hadriel Kaplan <hadriel.kap...@oracle.com> wrote:
>
> On Mar 21, 2014, at 5:31 PM, Guy Harris <g...@alum.mit.edu> wrote:
>
>> On Mar 21, 2014, at 2:18 PM, Hadriel Kaplan <hadriel.kap...@oracle.com>
>> wrote:
>>
>>> How long does the clang code analysis buldbot
>>
>> You mean the "beat the living hell out of Wireshark to look for bugs in many
>> different ways" buildbot? :-)
>
> What does it actually *do*?
The partial list in my previous message said:
scan-build, running the Clang static analyzer;
cov-build, building to prepare to hand Wireshark off to Coverity;
fuzz-menagerie, repeatedly running fuzzed capture files through TShark
to make sure it doesn't crash.
There's also a phase in which it runs capture files with random data ("random"
in the sense of "generated by a pseudo-random number generator") through TShark
as well.
> It looks like some of the crash bugs it submits are GUI-specific
Unlikely, given that it's testing TShark, not Wireshark, with the fuzzed
captures. See tools/fuzz-test.sh.
> and only occur way down the packet list, so presumably it actually loads the
> fuzzed files in wireshark
Nope. It really does run them through TShark; where has the buildbot filed a
bug in which *Wireshark* crashed?
> Does it also try some of the graphs/analyzers?
No, it doesn't test any of the Wireshark GUI.
> I ask because I can't repo a crash with bug 9887 by just viewing it,
Perhaps:
1) for some unknown reason, it crashes on Ubuntu (64-bit x86) but not
OS X (64-bit x86 in your case, as you are, as I remember, running a recent OS X
version);
2) you have different preferences set (I think the buildbot has an
empty set of preferences, so the defaults are used);
3) perhaps, for some reason, it crashes in TShark but doesn't crash in
Wireshark - for example, if you don't have any color filters set and don't have
a display filter, the path where it reads in the capture doesn't generate a
protocol tree, but I think the TShark test generates a protocol tree.
> but if I select Telephony->Voip Calls... boom goes the dynamite.
That might be a separate bug.
___________________________________________________________________________
Sent via: Wireshark-dev mailing list <wireshark-dev@wireshark.org>
Archives: http://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe
