Closed Bug 1767354 Opened 4 months ago Closed 3 months ago

Print out less information when leakcheck passes

Categories

(Testing :: General, task)

Default
task

Tracking

(firefox102 fixed)

RESOLVED FIXED
102 Branch
Tracking Status
firefox102 --- fixed

People

(Reporter: mccr8, Assigned: mccr8)

Details

Attachments

(1 file)

Whenever leakcheck.py processes a leak log, it prints out all of this for a single process:

INFO - leakcheck | Processing leak log file /tmp/tmp_97nr4s7.mozrunner/runtests_leaks_877_tab_pid19982.log
INFO - 
INFO - == BloatView: ALL (cumulative) LEAK AND BLOAT STATISTICS, tab process 19982
INFO - 
INFO -      |<----------------Class--------------->|<-----Bytes------>|<----Objects---->|
INFO -      |                                      | Per-Inst   Leaked|   Total      Rem|
INFO -    0 |TOTAL                                 |       52        0|   23153        0|
INFO - 
INFO - nsTraceRefcnt::DumpStatistics: 745 entries
INFO - TEST-PASS | leakcheck | tab no leaks detected!

Before Fission, there were only ever a few processes, so it wasn't a huge deal, but now a single test directory can have dozens or more processes.

Looking at a random WPT log, out of the 235,000 lines in the log, I'm estimating that about 100,000 are leakcheck related, which is pretty bad.

I think we should cut that down a bit in the passing case. It wouldn't be too hard to trim that down to this:

INFO - leakcheck | Processing leak log file /tmp/tmp_97nr4s7.mozrunner/runtests_leaks_877_tab_pid19982.log
INFO - TEST-PASS | leakcheck | tab no leaks detected!

That would save us 80,000 lines, which is something.

Mostly I find it a big pain scrolling back through page after page of useless output when I'm investigating a leak.

I did a try run to confirm that the output is unchanged when there's a leak. See the two Wd3 failures.

The passing output is as described in the previous comment. Arguably we could trim this down a bit further if there are no leaks in any content process or whatever (also maybe not dump the file name when we pass by default) but this at least helps for now.

This does change the behavior in the case of the weird failures, like negative leaks (we won't print the header information), but eh I think that's fine.

Pushed by amccreight@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/59ec9812f38f
Print out less information when leakcheck passes. r=ahal
Status: NEW → RESOLVED
Closed: 3 months ago
Resolution: --- → FIXED
Target Milestone: --- → 102 Branch
You need to log in before you can comment on or make changes to this bug.