I'd like to re-enable omni.ja compression on nightly that was disabled in bug 1231379 so we can measure the impact that uncompressed omni.jas have via telemetry, and also so that nightly matches aurora, beta and release.

Comment on attachment 8838054 [details] Bug 1340157: Re-enable compression for omni.ja files https://reviewboard.mozilla.org/r/113066/#review114554

Pushed by catlee@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/29f4d23a9e97 Re-enable compression for omni.ja files r=ted

this is causing talos to fail on linux64 pgo only: https://treeherder.mozilla.org/#/jobs?repo=autoland&group_state=expanded&filter-searchStr=talos%20other%20linux%20pgo&tochange=29f4d23a9e97c324b40c63c71eff04add5fd2a09&fromchange=22d650dc282e70b96de2fbaac84bf69a22cd3b54&selectedJob=78488722

Talos results showed: * increase in ts_paint (bad, and expected) * increase in sessionrestore time (bad, and expected) * decrease in Main_RSS (good, and expected) * decrease in main_startup_fileio (good, and unexpected) In contrast, data from Telemetry (looking at data since Feb 17th) seems to show: * decrease in SIMPLE_MEASURES_FIRSTPAINT (good!) https://mzl.la/2mmbzkD * decrease in SIMPLE_MEASURES_SESSIONRESTORED (good!) https://mzl.la/2mbwtXC * decrease in MEMORY_TOTAL (good!) https://mzl.la/2mml3wb I'm confused by the difference between Talos results and Telemetry. Given the unclear benefits here, and the fact that we didn't let this ride the trains before, I'm leaning towards leaving omni.jas compressed for now.

in many cases our machines for Talos are not representative of end user machines- but they do provide a good idea if things are good or bad. When we move 54 to Aurora, I will check back with this bug unless we resolve it as wontfix (from a talos regression standpoint)

(In reply to Joel Maher ( :jmaher) from comment #7) > When we move 54 to Aurora, I will check back with this bug unless we resolve > it as wontfix (from a talos regression standpoint) Aurora already has omni.ja compressed, because it never had it decompressed in the first place, so nothing will change on aurora.

Based on Telemetry, I'm going to leave the omni.ja files compressed.

Removing leave-open keyword from resolved bugs, per :sylvestre.

In bug 1362377 we decided to ship uncompressed omni.ja files in Firefox 55.