400842 - Remove memory leak debug data from output.log.

Status: RESOLVED FIXED

(NSS :: Test, defect, P2)

Description

[Slavomir Katuscak]

Memory leak checking reports output from leak checking tools to output.log and to separated files in memleak_logs directory. It's duplicity which is not desirable. 

After enabling PKIX testing, output.log became too big. We don't need stack details directly in output.log, it's enough to have them only in files in memleak_logs directory.

Comment 1

[Slavomir Katuscak]

Attachment: Patch.

Comment 2

[Nelson Bolyard (seldom reads bugmail)]

I believe the patch attached does what Slavo intended, but I am not sure that
we really want to make that change.  I don't agree that "it's enough to have 
them only in files in memleak_logs directory" unless the files in that 
directory (as a result of Tinderbox runs and/or nightly builds) are always 
readily available to NSS developers both inside and outside of Sun.  

Comments?

Comment 3

[Slavomir Katuscak]

We have one serious problem with big logfiles - after enabling PKIX there was log 2 times bigger then before (memleak.sh was tested 2 times) and mail server on Mozilla side just didn't accept it. 

Log from one test cycle on Linux/Valgrind has now 16.1 MB. When testing with PKIX (and now also with Shared DB) it increases much and mail server on Mozilla side has 25.6 MB limit for one email - that's also one of the reasons why I splitted Tinderbox testing to 4 cycles.

Before doing this, I got some returned mails:

   ----- The following addresses had permanent fatal errors -----
<[e-mail]>
    (reason: 552 5.2.3 Message size exceeds fixed maximum message size (25600000))

   ----- Transcript of session follows -----
... while talking to brmea-mail.sun.com.:
>>> MAIL From:<[e-mail]> SIZE=33184826
<<< 552 5.2.3 Message size exceeds fixed maximum message size (25600000)
554 5.0.0 Service unavailable

Comment 4

[Slavomir Katuscak]

Btw - my patch doesn't remove stacks from output.log completely, it prints there one line versions sorted with -u parameter (unique lines) - the same what is now in foundleaks file. 

Comment 5

[Nelson Bolyard (seldom reads bugmail)]

Comment on attachment 285897 [details] [diff] [review]
Patch.

OK, let's try it.

Comment 6

[Slavomir Katuscak]

Previous patch causes that Linux parser can't find the name of tested binary in Valgrind log file. I'm working on fix.

Comment 7

[Slavomir Katuscak]

Attachment: Patch to report binary name (used by Valgrind parser).

Fixed problem for comment 6. Previous patch removed message containing binary name from logs, this patch adds it back (in shorter format).

Comment 8

[Slavomir Katuscak]

Checking in memleak.sh;
/cvsroot/mozilla/security/nss/tests/memleak/memleak.sh,v  <--  memleak.sh
new revision: 1.21; previous revision: 1.20
done

Comment 9

[Slavomir Katuscak]

Checking in memleak.sh;
/cvsroot/mozilla/security/nss/tests/memleak/memleak.sh,v  <--  memleak.sh
new revision: 1.1.2.17; previous revision: 1.1.2.16
done