-
Notifications
You must be signed in to change notification settings - Fork 3.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[C++][CI] test-conda-cpp-valgrind fails on arrow-dataset-file-parquet-encryption-test #38304
Comments
@pitrou this seems to be a possible leak that should be fixed for the release, right? |
My understanding is that it's an initialization-related leak that is not easy to avoid properly. The reason it appeared in dataset is probably because dataset loads data from multiple threads, and OpenSSL can need extra care from the application to release per-thread memory As such, this leak is certainly benign and does not need to be addressed for the release. |
That said, I wonder which of the situations in that NOTES section is being hit, because it seems to me that none should apply here:
So it's a bit of a mystery why this occurs. Also cc @westonpace |
I'll try to play with this a bit more, but to reiterate: this isn't release-blocker. |
Thanks @pitrou ! |
Looking at the Valgrind output in more detail, it seems there are two different leaks, both of which are reported in a thread pool's worker thread:
|
…set-file-parquet-encryption-test If OpenSSL initializes itself from a non-main thread, it can fail deallocating all memory at shutdown. This is really a benign leak, but we don't want any spurious CI errors.
So, this memory is some static memory and forget to free them during shutdown? |
That's my understanding, yes. This is memory that seems to be allocated once when initializing OpenSSL (though |
…le-parquet-encryption-test (#38306) If OpenSSL initializes itself from a non-main thread, it can fail deallocating all memory at shutdown. This is really a benign leak, but we don't want any spurious CI errors. * Closes: #38304 Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Antoine Pitrou <antoine@python.org>
…set-file-parquet-encryption-test (apache#38306) If OpenSSL initializes itself from a non-main thread, it can fail deallocating all memory at shutdown. This is really a benign leak, but we don't want any spurious CI errors. * Closes: apache#38304 Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Antoine Pitrou <antoine@python.org>
…set-file-parquet-encryption-test (apache#38306) If OpenSSL initializes itself from a non-main thread, it can fail deallocating all memory at shutdown. This is really a benign leak, but we don't want any spurious CI errors. * Closes: apache#38304 Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Antoine Pitrou <antoine@python.org>
…set-file-parquet-encryption-test (apache#38306) If OpenSSL initializes itself from a non-main thread, it can fail deallocating all memory at shutdown. This is really a benign leak, but we don't want any spurious CI errors. * Closes: apache#38304 Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Antoine Pitrou <antoine@python.org>
Describe the bug, including details regarding any error messages, version, and platform.
The arrow-dataset-file-parquet-encryption-test seems to have an issue when running the valgrind jobs both on the maintenance branch and main:
The error:
Component(s)
C++, Continuous Integration
The text was updated successfully, but these errors were encountered: