Jeff Morriss
changed
bug 8431
| What |
Removed |
Added |
| CC |
|
jeff.morriss.ws@gmail.com
|
Comment # 25
on bug 8431
from Jeff Morriss
(In reply to comment #23)
> Created attachment 10236 [details]
> Add ddd debugger launch capabilities to test-fuzzed
>
> I added a -d option to launch ddd for the test-fuzzed replay script. As a
> non-C guru it was not clear to me that you need to
> 1. re-run with random environment variables to get the same failure as fuzz
> and
> 2. run everything from libtool
Hmmm, why should running 'ddd' or 'gdb' be done through the test-fuzzed script?
I admit I normally use that script just to get the core dump and run the
debugger on the resulting core file rather than running gdb on the process as
it experiences the problem. If that's commonly done then maybe it makes sense
to be able to launch the debugger from this script.
(I admit to run gdb I usually just type "Escape-/gdb" to search through my
shell's history. :-))
> Also, it would be nice if there were a way to turn off -O2 easily so the
> debugger is more effective. I manually hacked configure and did a make
> clean, but perhaps there is an easier way.
Usually I do something like this:
1) run test-fuzzed-cap on the offending PCAP file
2) debug the core file with gdb. If optimization is getting in the way I:
2.a) edit epan/dissectors/Makefile and take out "-O2" (there's only 1 place
that matters)
2.b) touch the dissector.c which caused the crash (so only it will be
rebuilt)
2.c) make -C epan
You are receiving this mail because:
- You are watching all bug changes.
