Open Bug 920366 Opened 6 years ago Updated 2 years ago

Clobber needed due to changes in memory reporting

Categories

(Firefox Build System :: General, defect)

defect
Not set

Tracking

(Not tracked)

People

(Reporter: philor, Unassigned)

References

(Blocks 1 open bug)

Details

See the Win opt in https://tbpl.mozilla.org/?tree=Mozilla-Inbound&rev=9195be8a50cb, the Win opt, ASan, and Linux64 debug in https://tbpl.mozilla.org/?tree=Mozilla-Inbound&showall=1&rev=eef35ea0291b, the Mac debug in https://tbpl.mozilla.org/?tree=Mozilla-Inbound&showall=1&rev=a87ff44842e6

Easiest way to tell you've reproed, after building debug 5cf98b6e4b18 and then 9195be8a50cb, ought to be to run xpcshell - reproing is a crash in test_TelemetryPing.js and test_TelemetryTimestamps.js.
gps, don't suppose you could look at this at some point? Would quite like to stamp out these various needs-clobbers, just due to the amount of infra load it causes us to waste when we mass force clobber all platforms on all trees every few days. Thank you :-)
Flags: needinfo?(gps)
I, uh, have no clue why this didn't work.

The offending changesets seem to be 8a8691a26012:9195be8a50cb. These changed a few .cpp and .idl files. From the crash log in the xpcshell tests, it appears the .cpp changes didn't get picked up. However, I looked through the Linux64 debug log (https://tbpl.mozilla.org/php/getParsedLog.php?id=28322271&tree=Mozilla-Inbound&full=1) and it seemed that the .xpt was regenerated and all the changed .cpp files were recompiled and linked. So I'm not sure what's going on here. Is there some kind of weird "cross-module" dependency in memory reporting land the build system doesn't know about?

njn: do you notice any weird behavior when modifying these files locally?
Depends on: 913260
Flags: needinfo?(n.nethercote)
Summary: Bug 913260 needed a clobber → Clobber needed due to changes in memory reporting
I built 8a8691a26012^1 and test_TelemetryPing.js didn't pass. So, I'm unable to reproduce this.

I could put more time into this. But I'd like to hear from njn first.
test_TelemetryPing.js doesn't work when run locally;  only when run on the test machines :(  If you comment out the line that fails, it works ok.

As for weird cross-module dependencies... xpcom/base/nsMemoryReporterManager.cpp now has a number of "registration" functions where other part of the browser (xpconnect, images, storage, dom) register callbacks.  See the DEFINE_REGISTER_DISTINGUISHED_AMOUNT macro.  Maybe that's the cause?
Flags: needinfo?(n.nethercote)
Flags: needinfo?(gps)
Blocks: clobber
Product: Core → Firefox Build System
You need to log in before you can comment on or make changes to this bug.