You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "WillAyd (via GitHub)" <gi...@apache.org> on 2025/03/10 20:50:11 UTC
[PR] fix(ci): Prevent flaky ASAN failures with Dremio [arrow-adbc]
WillAyd opened a new pull request, #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604
Per comment https://github.com/apache/arrow-adbc/pull/2546#issuecomment-2685799436 this might help fix the flaky CI failures with meson + ASAN
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Skip flaky ASAN failures in Meson [arrow-adbc]
Posted by "lidavidm (via GitHub)" <gi...@apache.org>.
lidavidm commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2833704069
Yeah, we do turn off leak checking and preload ASAN, but I found that I also needed to change the ASLR randomization bits to get ASAN to work consistently
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Prevent flaky ASAN failures with Dremio [arrow-adbc]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2825015231
After some more research, the patch that was applied to clang had already been in place with gcc since 2013. So I think I erred in assuming a newer version would change this. Need to go back to the drawing board...
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Prevent flaky ASAN failures with Dremio [arrow-adbc]
Posted by "lidavidm (via GitHub)" <gi...@apache.org>.
lidavidm commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2713116067
Seems we need to set more (maybe both CXX and CC have to be set?)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Skip flaky ASAN failures in Meson [arrow-adbc]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd merged PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Prevent flaky ASAN failures with Dremio [arrow-adbc]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2718159914
I asked the Meson team about this on their IRC channel, and they mentioned that while gcc uses shared linkage to the ASAN runtime, clang requires static linkage to ASAN. It seems like something is awry with the latter, and it _might_ have to do with how gtest is being used.
I will have to investigate that further, or maybe see if there's a newer version of gcc that can be used instead of clang that still fixes the issue
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Skip flaky ASAN failures in Meson [arrow-adbc]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2830436554
For now I've given up on trying to solve this, instead skipping the flightsql and bigquery tests in ASAN where I think the problem arises
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Skip flaky ASAN failures in Meson [arrow-adbc]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2830459316
I do wonder if this affects snowflake as well, since bigquery, flightsql and snowflake are all built using custom go targets...
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix(ci): Skip flaky ASAN failures in Meson [arrow-adbc]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd commented on PR #2604:
URL: https://github.com/apache/arrow-adbc/pull/2604#issuecomment-2833484098
From personal experience I've found ASAN challenging to use when launched via the Python interpreter. ASAN itself will detect some leaks from most version of CPython, so the only way I have gotten it to reliably run is to turn off leak detection while also making sure to LD_PRELOAD the ASAN library (this is required unless the Python interpreter itself was built with ASAN support)
```bash
ASAN_OPTIONS=detect_leaks=0 LD_PRELOAD=$(gcc -print-file-name=libasan.so)" python ...
```
Without that, I have seen that issue where you end up with an endless loop of ASAN_ERROR:DeadlySignal (or something to the effect) which is the error you are seeing right?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org