Well here's a good example. These two bugs are out-of-memory fuzz
failures that failed at the same time of day (one is master-1.12 and one
is master-1.10):
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11243
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11244
It certainly appears the machine is simply running low on memory while
testing.
On 05/30/2015 10:46 PM, Jeff Morriss wrote:
I ended up focusing on master-1.10 because there were so many bugs out
there. I guess that's just because most of the bugs against master were
already dup'd out!
I was wondering if maybe all the fuzz bots are running on the same
(virtual)machine and may be contending for memory (e.g., if two or more
of the buildbots run Valgrind at the same time then the box simply runs
out of memory).
On 05/30/2015 10:27 PM, Evan Huus wrote:
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11147 and its many
duplicates suggest a similar issue with the master buildbot.
On Sat, May 30, 2015 at 10:22 PM, Jeff Morriss
<jeff.morriss.ws@xxxxxxxxx> wrote:
There have been plenty of fuzz failures from the 1.10 branch in the
past few
months, including:
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11050
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11065
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11077
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11090
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11094
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11150
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=11239
All of the captures associated with those bugs run fine for me. The
last
one's error:
/home/wireshark/builders/wireshark-1.10-fuzz/fuzztest/build/.libs/lt-tshark:
error while loading shared libraries: libwireshark.so.3: failed to map
segment from shared object: Cannot allocate memory
really supports the theory that it's not the software that has the
problem
but the buildbot itself.
Is there something wrong with the box?
In the mean time I'll close all those bugs as WORKSFORME.
