Wireshark-dev: Re: [Wireshark-dev] [Bug 7060] Global variables - Buildbot crash output: fuzz-20
From: Evan Huus <[email protected]>
Date: Fri, 3 May 2013 11:43:27 -0400
(Taking this to -dev because it has nothing to do with the bug anymore)

On Fri, May 3, 2013 at 5:21 AM,  <[email protected]> wrote:
> Comment # 14 on bug 7060 from Guy Harris
>
> (In reply to comment #13)
>> (In reply to comment #11
> )
>> > (Not that this shouldn't be done
>> > - Global Variables Considered Harmful - but that we need to be careful
>> > when
>> > we do so.  Maybe I should try one of my big regression tests between the
>> > code without this change and the code with it.)
>>
>> Are these big regression tests sharable generally or on the wiki already
>> somewhere? Or are they not automated?
>
> The files are a combination of non-public capture I have, wireshark.org's
> "Wireshark menagerie" also used as source files for fuzz-testing (composed
> of
> captures harvested from bugs - Gerald, are they also harvested from mailing
> lists and/or sample captures on the Wiki?), and various old files from
> various
> sets of sample captures for other analyzers.
>
> I suspect most of the interesting problems show up with captures in the
> menagerie.

Is there
- an easy way to bulk-download parts of the menagerie?
- an easy way to bulk-add captures to the menagerie (short of creating
a dummy bug)?

> The scripts are fairly dumb scripts that:
>
>     run "before" and "after" versions of "tshark -V", piping them into "cmp"
> and, if "cmp" says they're not the same, runs those versions again, with the
> standard output sent to "/tmp/before" and "/tmp/after", and diffs them;
>
>     does a recursive "find" and runs the aforementioned script on the
> capture
> files it finds.
>
> The first of those scripts could use a little generalization - right now,
> every
> time I have a new "wireshark.XXX" directory and a "wireshark.XXX-baseline"
> directory, I copy the script and tweak it to look for
> ../wireshark.XXX-baseline/tshark" and compare its output with that of
> "tshark".
>
> (They also will, for a capture file named foo, look for a foo.options file
> and,
> if it finds it, sticks its contents into the tshark command lines, to set
> options with "-o".)

The scripts might be worth dumping in tools/ at some point. The
options feature, in particular, sounds like it might be worth adding
to the fuzz scripts, since that would allow us to fuzz much more
interesting things. Whoever does that for GSoC might look into this.

> The results aren't available on the Web - the test machine is a laptop and
> is
> clamshelled when I'm not using it, so it's not going to be much of a server.
>
> ________________________________
> You are receiving this mail because:
>
> You are on the CC list for the bug.