38289 - Mac daily verification build fails

Status: VERIFIED FIXED

(SeaMonkey :: Build Config, defect, P2)

Description

[Jon Granrose]

The mac daily verification build is failing to build.  I'll attach the macperl 
and buildnglayout logs to the bug.

This is a blocker for the mac.

Comment 1

[Jon Granrose]

reassigning to jj instead of cls.

Comment 2

[Jon Granrose]

Attachment: MacPerl Log file

Comment 3

[Jon Granrose]

Attachment: Build NG Layout log file

Comment 4

[leger]

Putting on dogfood+ radar.

Comment 5

[Jon Granrose]

By doing a "cvs checkout" in the mozilla/mailnews directory followed by a build 
where I didn't delete the existing tree (depend/clobber in mac automation), the 
mozilla build succeeded and the commercial build is still building.

However, I think we still have to discover the root cause of the problem that it 
isn't pulling and/or building everything in mozilla/mailnews.

removing blocker status since we apparently have a workaround.

Comment 6

[J.J. Enser]

i just ran the pull section of buildcentral (name of the main verif/tinderbox 
script) and witnessed the same behavior as this morning: no error, no timeout, a 
few warnings (attached screen shot for the record), but mailnews is incomplete 
(see second attachment)
Obviously Tinderboxen don't experience this problem since they're all green at 
the moment.

can anybody else try to run pullnglayout.pl on a fresh tree and see if this can 
be reproduced on a different machine ? I'm running out of ideas.

Comment 7

[J.J. Enser]

Attachment: CVS pull warnings

Comment 8

[J.J. Enser]

Attachment: mailnews dir content ('public' missing)

Comment 9

[J.J. Enser]

fixed. this is a MacCVS low memory situation. When this happens, MacCVS simply 
stopped pulling "SeaMonleyAll" module before its end. No timeout returned to 
MacPerl, no error in the message window, no alert... I had to trace the build 
script to see this with my own eyes and notice that the free heap was around 50K 
at the time.

Increasing MacCVS memory partition from 5 to 8 MB fixed the problem and the 
entire tree now gets pulled without a glitch.

This is still a mystery to me since all Tinderboxen are configured identically (5 
MB for MacCVS, 256 MB machines) and don't have this problem.

maccvs team: Should I log a separate "MacCVS bug"? Would switching to 2.5 help ?

Comment 10

[Jon Granrose]

no failure yesterday or today.  verified.