[Evan Huus <eapache@xxxxxxxxx>]

However, looking more closely at the times: there is *no way* given the size of our menagerie, that it should be valgrinding the entire thing in under half an hour (http://buildbot.wireshark.org/trunk/builders/Clang-Code-Analysis/builds/2061).

I added another tweak in 51300 that might give a clue what's going on. Gerald, is it possible that valgrind isn't in the PATH or doesn't have execute permission for the fuzz-bot's environment/user? Something like that might explain the 1/2-hour time, since then valgrind would never be run at all...

On Sun, Aug 11, 2013 at 1:33 PM, Evan Huus <eapache@xxxxxxxxx> wrote:

Well, at least it seems to be running correctly with and without tree now. Part of r50895 probably fixed it by accident somehow...

On Thu, Jul 25, 2013 at 8:37 PM, Evan Huus <eapache@xxxxxxxxx> wrote:

I'm coming to the conclusion that it isn't working at all, even apart
from the weird only-one-set-of-arguments bug that we never tracked
down. I just grabbed a chunk of the menagerie and ran fuzz-test -g on
it myself and I'm finding way too many errors for this to be just me
getting lucky.

This suggests, however, that if we were to fix it we would be flooded
in fuzz failures...

Evan