It makes the shell compilation much slower because of lalrpop. As for Cranelift, we can make it easy to disable compilation for it.

the dependency for lalrpop is about to be removed in bug 1520001.

https://hg.mozilla.org/mozilla-central/rev/2d4f87a0212e

Seems like this improved our performance on Android! \0/

== Change summary for alert #19406 (as of Fri, 15 Feb 2019 03:07:41 GMT) ==

Improvements:

3% raptor-speedometer-geckoview android-hw-p2-8-0-arm7-api-16 opt 22.37 -> 23.00
2% raptor-speedometer-geckoview android-hw-p2-8-0-android-aarch64 opt 20.72 -> 21.20

For up to date results, see: https://treeherder.mozilla.org/perf.html#/alerts?id=19406

(In reply to Ionuț Goldan [:igoldan], Performance Sheriffing from comment #5)

Seems like this improved our performance on Android! \0/

== Change summary for alert #19406 (as of Fri, 15 Feb 2019 03:07:41 GMT) ==

Improvements:

3% raptor-speedometer-geckoview android-hw-p2-8-0-arm7-api-16 opt 22.37 -> 23.00
2% raptor-speedometer-geckoview android-hw-p2-8-0-android-aarch64 opt 20.72 -> 21.20

For up to date results, see: https://treeherder.mozilla.org/perf.html#/alerts?id=19406

Well, there's no apparent reason why it did so... Is there a way to double-check it was indeed related to this patch?

(In reply to Benjamin Bouvier [:bbouvier] from comment #6)

(In reply to Ionuț Goldan [:igoldan], Performance Sheriffing from comment #5)

Seems like this improved our performance on Android! \0/

== Change summary for alert #19406 (as of Fri, 15 Feb 2019 03:07:41 GMT) ==

Improvements:

3% raptor-speedometer-geckoview android-hw-p2-8-0-arm7-api-16 opt 22.37 -> 23.00
2% raptor-speedometer-geckoview android-hw-p2-8-0-android-aarch64 opt 20.72 -> 21.20

For up to date results, see: https://treeherder.mozilla.org/perf.html#/alerts?id=19406

Well, there's no apparent reason why it did so... Is there a way to double-check it was indeed related to this patch?

Sure there is. I will need to test this on Try by comparing a baseline with a backout. Keeping the ni? as a reminder.

I retriggered this a couple of times and seems like the pattern holds. If the improvements shouldn't relate to this bug, then maybe they're more related to bug 1524337? As this patch included it also.

Thanks Ionuț for checking; I've exposed a theory in bug 1524337 where this same strange thing happened too.