New fuzz harness + refactoring into fuzz/ #2273
Conversation
|
Once #2273 will be merged, this google/oss-fuzz#6607 should succeed. |
|
I checked the output of all 3 failed tests (fastdds_aarch64, fastdds_mac, and fastdds_windows), but none of the failures seem caused by this commit. Can someone take a look? |
There was a problem hiding this comment.
Failed tests are not related to the changes on this PR.
Apart from the comments below, please remove uncrustify.cfg (we have it on a different repo)
|
|
||
| return true; | ||
|
|
||
| #ifdef EMULATE_WRITER |
There was a problem hiding this comment.
What's the purpose of this code? It seems the created writer is not being used after it is created ...
There was a problem hiding this comment.
We can leave it or we can drop it. My intention was to leave it there for future testing, but now is too unstable when running in a fuzzing environment because threading makes fuzzing runs non-deterministic. I don't know Fast-DDS that well to be able to bypass threading and leave the rest untouched, but if you can see a way to integrate a "full" RTPS system around the simple processCDRMsg, I think this could increase coverage and possibly find more bugs.
There was a problem hiding this comment.
I think that could be achieved create a custom transport. You will need to implement TransportInterface and TransportDescriptorInterface.
You would then need to set useBuiltinTransports = false on RTPSParticipantAttributes, and also add the new transport descriptor on userTransports.
There was a problem hiding this comment.
Understood. Is this a good reference example for this use case? https://github.com/eProsima/Fast-DDS/tree/master/examples/C%2B%2B/UserDefinedTransportExample
There was a problem hiding this comment.
Yeah. In particular this lines are the ones related with setting the custom transport.
There was a problem hiding this comment.
OK. I'm eyeing at the SharedMemTransportDescriptor as a good starting point for a custom transport that doesn't involve any threading, but I won't be able to integrate it soon, unless you think that SharedMemTransportDescriptor is ready for the purpose.
In any case, since the integration of this harness is blocking the integration of the new XML harness, I think I'll just remove this extended feature in this PR, so it can hopefully be integrated, and postpone it to a separate PR.
- Appended `uncrustify.cfg` pattern into `.gitignore` - Moved all harnesses into `fuzz/` - Refactored make files I sent a PR to google/oss-fuzz with these commits to conclude integration of refactored fuzzers. Will keep an eye on the build process and fix it if something will go wrong. https://github.com/google/oss-fuzz/compare/master...phretor:fast-dds/refactor?expand=1 Signed-off-by: Federico Maggi <fede@maggi.cc>
cba765e to
1a14650
Compare
|
Updated as requested. Failing jobs should be unrelated to this PR. |
Signed-off-by: Federico Maggi <federico@maggi.cc>
Whoops! Too many git resets and I forgot about this :-) should be removed now. |
Signed-off-by: Federico Maggi <federico@maggi.cc>
I'll send a PR to google/oss-fuzz with these commits to conclude integration of refactored fuzzers. Will keep an eye on the build process and fix it if something will go wrong.
https://github.com/google/oss-fuzz/compare/master...phretor:fast-dds/refactor?expand=1
Signed-off-by: Federico Maggi fede@maggi.cc