investigate flaky parallel/test-crypto-dh-leak #21076
Comments
Trott
added
crypto
|
By the way, that was the fourth time in a row that happened on https://ci.nodejs.org/job/node-test-commit-linux-containered/nodes=ubuntu1604_sharedlibs_debug_x64/4846/console |
|
@nodejs/crypto @nodejs/build @nodejs/testing |
|
This comment above the assertion that's failing suggests that this may be inherently unreliable? // RSS should stay the same, ceteris paribus, but allow for
// some slop because V8 mallocs memory during execution. |
|
@Trott if definitely doesn't fail on MacOS. I'll try to look into it. |
Shouldn't this comment be improved? Of all things: why is there latin in there? |
|
Because being a programmer doesn't mean you can't be well-versed in the liberal arts too. Ceteris paribus means "all other things being equal" but with fewer words. Is the |
|
We might be able to circumvent that by massively increasing the number of iterations and slightly increasing the amount of permitted RSS difference. |
|
Other options: We could check Or @tniessen is suggesting we could remove the test entirely as it was for a very specific regression. Thoughts? @bnoordhuis |
|
On my maschine, the debug build allocates 4.8MB of the 5MB limit, but that value does not increase when I increase the number of iterations, so #21076 (comment) could work. |
It transpires that the extra bookkeeping in debug builds sometimes makes the increase in RSS go _just_ over the 5 MB limit, by fewer than 100 kB. Double the limit so we hopefully don't run into it any time again soon. The memory leak it tests for was one where RSS grew by hundreds of megabytes over the lifetime of the test; 5 vs. 10 MB is insignificant. Fixes: nodejs#21076
Refs: nodejs#21076 PR-URL: nodejs#21080 Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Rich Trott <[email protected]> Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Сковорода Никита Андреевич <[email protected]> Reviewed-By: Yuta Hiroto <[email protected]> Reviewed-By: James M Snell <[email protected]> Reviewed-By: Tiancheng "Timothy" Gu <[email protected]> Reviewed-By: Luigi Pinca <[email protected]>
Refs: #21076 PR-URL: #21080 Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Rich Trott <[email protected]> Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Сковорода Никита Андреевич <[email protected]> Reviewed-By: Yuta Hiroto <[email protected]> Reviewed-By: James M Snell <[email protected]> Reviewed-By: Tiancheng "Timothy" Gu <[email protected]> Reviewed-By: Luigi Pinca <[email protected]>
It transpires that the extra bookkeeping in debug builds sometimes makes the increase in RSS go _just_ over the 5 MB limit, by fewer than 100 kB. Double the limit so we hopefully don't run into it any time again soon. The memory leak it tests for was one where RSS grew by hundreds of megabytes over the lifetime of the test; 5 vs. 10 MB is insignificant. Fixes: #21076 PR-URL: #21080 Refs: #21076 Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Rich Trott <[email protected]> Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Сковорода Никита Андреевич <[email protected]> Reviewed-By: Yuta Hiroto <[email protected]> Reviewed-By: James M Snell <[email protected]> Reviewed-By: Tiancheng "Timothy" Gu <[email protected]> Reviewed-By: Luigi Pinca <[email protected]>
https://ci.nodejs.org/job/node-test-commit-linux-containered/4846/nodes=ubuntu1604_sharedlibs_debug_x64/console
The text was updated successfully, but these errors were encountered: