Fix ASAN issues with std::function usage
#4673
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
david-cortes
requested review from
btrotta,
chivee,
guolinke and
shiyu1994
as code owners
|
Thank you so so so much for the quick response helping us with these issues from CRAN!!! I've created #4674 to track the discussion around this issue, and so we can reference it in the future. I think these changes look good, but I'd like to try to see if I can replicate the errors CRAN is reporting by following https://www.stats.ox.ac.uk/pub/bdr/memtests/README.txt, then test this PR with that setup. I'm going to attempt that tonight or tomorrow. |
|
I was able to reproduce the ASAN errors CRAN reported tonight. I tried those same steps on this branch to see if the changes here resolve the issue, and it seems that they don't. code I ran to test this branch with ASANgit clone \
--recursive \
--branch fptr \
git@github.com:david-cortes/LightGBM.git \
david-cortes-lgb
cd david-cortes-lgb
docker run \
--rm \
-v $(pwd):/usr/LightGBM \
--workdir /usr/LightGBM \
-it \
wch1/r-debug:latest \
/bin/bash
export ASAN_OPTIONS="detect_leaks=0:detect_odr_violation=0"
export UBSAN_OPTIONS="print_stacktrace=1"
export RJAVA_JVM_STACK_WORKAROUND=0
export RGL_USE_NULL=true
export R_DONT_USE_TK=true
RDscriptsan -e "install.packages(c('R6', 'data.table', 'jsonlite', 'Matrix', 'testthat'), repos = 'https://cran.r-project.org', Ncpus = parallel::detectCores())"
sh build-cran-package.sh
RDsan CMD INSTALL lightgbm_*.tar.gz
cd R-package/tests
RDscriptsan testthat.R >> tests.log 2>&1
# check that tests passed
echo "test exit code: $?"
tail -300 ./tests.logI see the same issue in logs as CRAN noted:
full logsCould you take a look at the code I used above and see if you see any mistakes? If not, then this change probably does not fix the issue CRAN has noted. |
|
That's interesting - just noticed that CRAN shows two different errors, and the one I was thinking about was this: Perhaps there's another problem when it is not assigned a null pointer, but from a look at the trace, I'd venture to guess it could be a bug in the compiler header. Will investigate later. |
|
Did a bit more investigation:
|
david-cortes
changed the title
std::function usage
|
Thank you so much! Let's see what the CI jobs say about whether this change would break other things. I can also test with the |
|
closing and re-opening to retrigger CI, since we've recently fixed some configuration issues with Appveyor and Azure Pipelines jobs (#4675 (comment)). |
jameslamb
approved these changes
Oct 15, 2021
There was a problem hiding this comment.
I just tested this locally with the following code that seems to successfully reproduce CRAN's gcc-ASAN and clang-ASAN tests (as described in #4674 (comment) and in progress at #4678).
testing code (click me)
git clone \
--recursive \
--branch fptr \
git@github.com:david-cortes/LightGBM.git \
david-cortes-lgb
cd david-cortes-lgb
#
# options for CUSTOMIZATION:
#
# - "san" = gcc-ASAN
# - "csan" = clang-ASAN
#
docker run \
--rm \
-v $(pwd):/usr/LightGBM \
--workdir /usr/LightGBM \
--env CUSTOMIZATION=csan \
-it \
wch1/r-debug:latest \
/bin/bash
RDscript${CUSTOMIZATION} \
-e "install.packages(c('R6', 'data.table', 'jsonlite', 'Matrix', 'testthat'), repos = 'https://cran.r-project.org', Ncpus = parallel::detectCores())"
sh build-cran-package.sh
RD${CUSTOMIZATION} \
CMD INSTALL lightgbm_*.tar.gz
cd R-package/tests
rm -f ./tests.log
RDscript${CUSTOMIZATION} testthat.R >> tests.log 2>&1
# check that tests passed
echo "test exit code: $?"
tail -300 ./tests.logNo issues were raised by those tests as of the current state of this branch 🎉
So I'm confident that when {lightgbm} 3.3.1 is submitted, this change will resolve the errors CRAN has noted in clang-ASAN and gcc-ASAN checks.
Thanks so much for the help, @david-cortes !
Unfortunately, we are not testing CLI version at CI. |
jameslamb
added a commit
that referenced
this pull request
Oct 22, 2021
This reverts commit 13ed38c.
jameslamb
added a commit
that referenced
this pull request
Oct 24, 2021
This reverts commit 24c275b.
StrikerRUS
added a commit
that referenced
this pull request
Oct 26, 2021
…ixes #4674) (#4678) * add jobs mimicking CRAN gcc-ASAN and clang-ASAN * comment out CI * fix redirection * remove unnecessary echo * Revert "comment out CI" This reverts commit 899fbb4. * remove redundant env variables and update README * remove inaccurate comment * change test title * Revert "Fix ASAN issues with `std::function` usage (#4673)" This reverts commit 13ed38c. * Revert "Revert "Fix ASAN issues with `std::function` usage (#4673)"" This reverts commit 24c275b. * revert unnecessary change in config order * Apply suggestions from code review Co-authored-by: Nikita Titov <nekit94-08@mail.ru> Co-authored-by: Nikita Titov <nekit94-08@mail.ru>
|
This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The C API uses a C++
std::functionwhich gets compared against anullptrto check if it assigned, which is not correct. This PR makes small changes to make the usage more idiomatic C++.I think this should fix the (or one of the) issue(s) highlighted by CRAN: #4633
But I'm not sure if this is all that was wrong or if there are more instances of usages like these, or if this will affect anything else other than the dataset construction in the C API.