1258968 - NSS memleak tests are broken after the removal of SSL v2 code

Status: RESOLVED FIXED

(NSS :: Test, defect)

Description

[Franziskus Kiefer [:franziskus]]

NSS memleak builbots are always failing. This bug is to track investigation into this.

Comment 1

[Kai Engert [:KaiE:]]

FYI, one of the memleak machines is inside the Mozilla network. If you'd like access, please attach your ssh public key.

Comment 2

[Kai Engert [:KaiE:]]

This is a consequence of the removal of SSL v2 code, bug 1228555.

selfserv doesn't work with the cipher suite list that the memleak.sh script attempts to use.

memleak.sh needs to be changed similar to 
https://hg.mozilla.org/projects/nss/diff/0c19072b8537/tests/ssl/ssl.sh#l89

Given that this is preventing us from having memory leak tests currently, I think it should be a blocker for the 3.24 release.

Comment 3

[Tim Taubert [:ttaubert] (inactive)]

Attachment: 0001-Bug-1258968-Fix-memleak-tests-after-removing-SSL-2-c.patch

Comment 4

[Kai Engert [:KaiE:]]

Comment on attachment 8742333 [details] [diff] [review]
0001-Bug-1258968-Fix-memleak-tests-after-removing-SSL-2-c.patch

Thanks for the patch, let's check it in and see if it fixes it, r=kaie

Comment 5

[Tim Taubert [:ttaubert] (inactive)]

https://hg.mozilla.org/projects/nss/rev/933ae21a1370

Comment 6

[Franziskus Kiefer [:franziskus]]

Looks like there are still problems on the debug builds [1]. It's always a timeout so probably unrelated to the SSLv2 code, but we should investigate what's going on. Maybe we can increase the timeout?

[1] https://bot.nss-crypto.org:8011/builders/8-linux-x64-DBG-memleak

Comment 7

[Tim Taubert [:ttaubert] (inactive)]

Yeah, increasing the timeout seems like a good first step. I can imagine a debug memleak build takes quite some time.

Comment 8

[Kai Engert [:KaiE:]]

Looks like the memleak builds are good now.

Comment 9

[Liz Henry (:lizzard) (relman/hg->git project)]

Is this fixed on all branches? It is still marked as affected in 48. Thanks!

Comment 10

[Tim Taubert [:ttaubert] (inactive)]

(In reply to Liz Henry (:lizzard) (needinfo? me) from comment #9)
> Is this fixed on all branches? It is still marked as affected in 48. Thanks!

As 48 is running with 3.24 I *think* that it should have the fix. It's fixing a test script only though so it doesn't affect Firefox directly.