Uh oh!
There was an error while loading. Please reload this page.
ARROW-12571: [R][CI] Run nightly R with valgrind - #10237
Conversation
jonkeane
commented
May 3, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: 0e1d0e219c3e05b52ad180bd626a499138b577d6 Submitted crossbow builds: ursacomputing/crossbow @ actions-377
|
There was a problem hiding this comment.
You should use valgrind --error-exitcode=1 instead of grepping.
There was a problem hiding this comment.
We should use a Valgrind suppressions file instead of grepping:
https://stackoverflow.com/questions/17159578/generating-suppressions-for-memory-leaks
There was a problem hiding this comment.
Why these changes? If they have a specific purpose, there should be a comment explaining them.
There was a problem hiding this comment.
Isn't this dangerous if there are CI secrets in environment variables? Why is this needed?
There was a problem hiding this comment.
I was using it to diagnose if the ARROW_RUNTIME_SIMD_LEVEL var was being passed — secrets should be redacted even with something like printenv, but I've removed it regardless.
pitrou
commented
May 4, 2021
I'm trying this out and I see that it takes a lot of time without outputting anything: Can the above command output progress while running the test suite? |
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: 041f4095fb71adfbfb5d504ff83da04c07cb4352 Submitted crossbow builds: ursacomputing/crossbow @ actions-379
|
Revision: 041f4095fb71adfbfb5d504ff83da04c07cb4352 Submitted crossbow builds: ursacomputing/crossbow @ actions-380
|
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: 4ede120d38854d6bbf7556450ed15e8345b59fcd Submitted crossbow builds: ursacomputing/crossbow @ actions-381
|
There was a problem hiding this comment.
Could also do something like (of course confirming the right way to determine a devel version). Could also add the devel check to the skip_on_linux_cran (and call it skip_on_valgrind()) if you wanted to make that narrower.
| # expect_identical(as.numeric(sum(na)), sum(floats)) | |
| expect_identical(as.numeric(sum(na)), NA_real_) | |
| if (!grepl("devel", R.version.string)) { | |
| # Valgrind on R-devel confuses NaN and NA_real_ | |
| expect_identical(as.numeric(sum(na)), sum(floats)) | |
| } |
There was a problem hiding this comment.
Yeah, I'll add the devel check to skip_on_linux_cran() (and probably rename it too, though I'm a bit hesitant to make it so obvious that we are skipping the problematic tests — is that something we should worry about?).
As for these specific tests, looking at it now I wonder if we have benefit from expect_identical(as.numeric(sum(na)), sum(floats)) versus expect_identical(as.numeric(sum(na)), NA_real_)? Is there a circumstance where we expect sum(floats) to change behavior and we would want to match it? I find expect_identical(as.numeric(sum(na)), NA_real_) to be easier to read at a glance and see exactly what it's testing/expecting.
Uh oh!
There was an error while loading. Please reload this page.
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: 7f29abc988154513ad8f43a789b3399c3dfd971c Submitted crossbow builds: ursacomputing/crossbow @ actions-384
|
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind (with mimalloc as default) |
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: a94cff3c451e708e387d27dd89e9832b2eedbfea Submitted crossbow builds: ursacomputing/crossbow @ actions-385
|
nealrichardson
left a comment
There was a problem hiding this comment.
Thanks a lot for taking this on.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
| ARROW_USER_SIMD_LEVEL: "AVX2" | |
| ARROW_RUNTIME_SIMD_LEVEL: "AVX2" |
There was a problem hiding this comment.
Have you confirmed that this job (in the script's final form) correctly fails if there's a valgrind failure? You could try this by making skip_on_valgrind() no-op and see if it fails as expected, then restore the skips.
There was a problem hiding this comment.
Yes it does, though I will re-confirm that once we are close to merging
7886dcb to
5a7493eCompare5a7493e to
3ab8871Comparejonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: 3ab8871 Submitted crossbow builds: ursacomputing/crossbow @ actions-389
|
jonkeane
commented
May 4, 2021
If ^^^ doesn't pass, there was a merge/squash error |
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
jonkeane
commented
May 4, 2021
We don't expect ^^^ to pass, it is intentionally broken |
Revision: 324786f Submitted crossbow builds: ursacomputing/crossbow @ actions-390
|
e1a8bd3 to
3bb94e6Comparejonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
Revision: 3bb94e6 Submitted crossbow builds: ursacomputing/crossbow @ actions-391
|
jonkeane
commented
May 4, 2021
We don't expect ^^^ to pass, it is intentionally broken |
jonkeane
commented
May 4, 2021
@github-actions crossbow submit test-r-linux-valgrind |
jonkeane
commented
May 4, 2021
And ^^^ should pass now that we reverted our intentional failures. |
Revision: 5812a19 Submitted crossbow builds: ursacomputing/crossbow @ actions-392
|
Closesapache#10237 from jonkeane/ARROW-12571-valgrindCI-patch Authored-by: Jonathan Keane <jkeane@gmail.com> Signed-off-by: Jonathan Keane <jkeane@gmail.com>
Closesapache#10237 from jonkeane/ARROW-12571-valgrindCI-patch Authored-by: Jonathan Keane <jkeane@gmail.com> Signed-off-by: Jonathan Keane <jkeane@gmail.com>
No description provided.