Migrates and briefly documents the UBet analysis code for public reference in the LangSec slides. #6552
Conversation
kaoudis
force-pushed
the
kkaoudis/migrate-ubet-python
branch
from
0658e67
to
7fd0e28
Compare
…parent dockerfile is now different, as also in #6551
…want (and so that I can use it when I need it later)
… for analysis if desired
|
|
||
| RUN cp modules/c++/nitf/show_nitf++ nitro_Release | ||
|
|
||
| RUN polytracker instrument-targets \ |
There was a problem hiding this comment.
We will need to add --cflog once #6553 is merged. Additional flag needed to not change existing behaviour.
There was a problem hiding this comment.
We need to add it here as the Dockerfile.nitro does not exist on #6553. Either we do it now, or we add it as a follow-up PR. We could also merge all our changes into a separate branch and then make the final changes there. But, I think merging #6553 first, then just adding the --cflog here and merging makes the most sense to me. Ideally this wouldn't be dependent on #6553 to be merged, in practice it is hard to avoid.
|
|
||
| RUN polytracker build cmake --build . -j$((`nproc`+1)) --clean-first --target show_nitf++ --config Debug | ||
| RUN cp modules/c++/nitf/show_nitf++ nitro_Debug | ||
| RUN polytracker instrument-targets \ |
There was a problem hiding this comment.
We will need to add --cflog once #6553 is merged. Additional flag needed to not change existing behaviour.
|
|
||
| RUN pip install cxxfilt | ||
|
|
||
| RUN git clone https://github.com/mdaus/nitro.git |
There was a problem hiding this comment.
I suggest we pin to the specific version of Nitro that we know works, just in case something upstream breaks our instrumentation later on.
This commit has I chose is from mid-March, but I didn't test it. @kaoudis do you think we need an earlier version?
| RUN git clone https://github.com/mdaus/nitro.git | |
| RUN git clone https://github.com/mdaus/nitro.git | |
| WORKDIR /polytracker/the_klondike/nitro | |
| RUN git checkout 342f552768e249e86df702062ff3f60ea1ec813a |
There was a problem hiding this comment.
I think committing a version that matches the version when we were working on this is probably best for now until I/we get a chance to add more analysis stuff here and start working on this again!
Looks like some recent Nitro commits are just merging changes from coda-oss, which may perhaps be used more inside Maxar (Nitro is the only open source use I am currently aware of) and coda-oss does seem to introduce a good chunk of Nitro's variability. I think pinning to b39ccc4c07e84e6c05cecb9ae24143373a3ed8e2 could be better since that would be the one that we were looking at when we did the work. What do you think?
There was a problem hiding this comment.
Yes! Let's pin the version you are suggesting @kaoudis! That's the one we have a working result from.
…eeds things up - follow the build job
…s to complete instead of hanging forever for right this second
❗ caution: depends on #6553 and will not accurately reproduce our results without merging that code first / too.