9MB GeckoView size increase
Categories
(Firefox Build System :: General, defect, P3)
Tracking
(firefox-esr68 unaffected, firefox-esr78 wontfix, firefox79 wontfix, firefox80 wontfix, firefox81 fixed)
Tracking | Status | |
---|---|---|
firefox-esr68 | --- | unaffected |
firefox-esr78 | --- | wontfix |
firefox79 | --- | wontfix |
firefox80 | --- | wontfix |
firefox81 | --- | fixed |
People
(Reporter: agi, Assigned: glandium)
References
(Regression)
Details
(Keywords: regression, Whiteboard: [geckoview])
Attachments
(1 file)
(deleted),
text/x-phabricator-request
|
Details |
The GVE APK (and all the other GV products including Fenix) increase from ~68MB to ~78MB This seems to have started with Bug 1640578? :glandium can you look at this?
Reporter | ||
Updated•4 years ago
|
Comment 1•4 years ago
|
||
I think this is essentially a dup of bug 1644553 -- or at least will be addressed by the patch in that bug.
Reporter | ||
Updated•4 years ago
|
Updated•4 years ago
|
Comment 2•4 years ago
|
||
S3/P3 on a 18% regression on libxul size for mobile apps? How about reconsidering that?
Before this change libxul size 46,056,749 bytes after this regression libxul size 54,248,458 bytes.
Comment 3•4 years ago
|
||
I don't think the condescension is appropriate or constructive.
We're waiting on a NI from someone who isn't in an American timezone. The bug will be resolved.
Assignee | ||
Comment 4•4 years ago
|
||
(In reply to Nathan Froyd [:froydnj] from comment #1)
I think this is essentially a dup of bug 1644553 -- or at least will be addressed by the patch in that bug.
Wrong bug?
Assignee | ||
Comment 5•4 years ago
|
||
It appears we were stripping local symbols in libraries before bug 1640578, which was actually an oversight from bug 818418: it wasn't actually intended that Android would be excluded. We do strip more fully when profiling is not enabled. Profiling is only enabled on central, not beta/release/esr. So unless it turns into a regression in beta, I'd call this wontfix.
Updated•4 years ago
|
Reporter | ||
Comment 6•4 years ago
|
||
(In reply to Mike Hommey [:glandium] from comment #5)
It appears we were stripping local symbols in libraries before bug 1640578, which was actually an oversight from bug 818418: it wasn't actually intended that Android would be excluded. We do strip more fully when profiling is not enabled. Profiling is only enabled on central, not beta/release/esr. So unless it turns into a regression in beta, I'd call this wontfix.
Hey glandium, thanks for looking at this. Why do we have symbols on Android Nightlies? I thought we didn't use them for anything.
Comment 7•4 years ago
|
||
(In reply to Mike Hommey [:glandium] from comment #4)
(In reply to Nathan Froyd [:froydnj] from comment #1)
I think this is essentially a dup of bug 1644553 -- or at least will be addressed by the patch in that bug.
Wrong bug?
I think I somehow assumed that profile generation had been turned on for artifact builds and we were therefore grabbing the wrong libxul. But that's not even the scenario described by the bug, so chalk that one up to my own confusion. :)
Assignee | ||
Comment 8•4 years ago
|
||
(In reply to Agi Sferro | :agi | ⏰ PST | he/him from comment #6)
(In reply to Mike Hommey [:glandium] from comment #5)
It appears we were stripping local symbols in libraries before bug 1640578, which was actually an oversight from bug 818418: it wasn't actually intended that Android would be excluded. We do strip more fully when profiling is not enabled. Profiling is only enabled on central, not beta/release/esr. So unless it turns into a regression in beta, I'd call this wontfix.
Hey glandium, thanks for looking at this. Why do we have symbols on Android Nightlies? I thought we didn't use them for anything.
They're useful for, among other things, the profiler.
Reporter | ||
Comment 9•4 years ago
|
||
(In reply to Mike Hommey [:glandium] from comment #8)
They're useful for, among other things, the profiler.
Do the symbols have to be in the device? could we make desktop download the symbols when using the profiler instead? 9MB of extra APK size (even if only for the nightly browser) is not something we should take lightly, especially for a developer-only feature like the profiler.
Reporter | ||
Updated•4 years ago
|
Reporter | ||
Comment 10•4 years ago
|
||
Reopening until Comment 9 gets addressed.
Assignee | ||
Comment 11•4 years ago
|
||
Upon reflection, we probably don't need the extra symbols anymore. Back
when we reduced the amount of stripping when enabling profiling, the
profiler wasn't able to download symbols on its own. It now is able to
do so, on all platforms.
As the stripping happens at packaging time, this doesn't change anything
for mach run on local builds.
Updated•4 years ago
|
Comment 12•4 years ago
|
||
Comment 13•4 years ago
|
||
bugherder |
Comment 14•4 years ago
|
||
IIUC, this doesn't affect the builds we actually end up shipping and therefore doesn't need backporting. Please let me know if I'm misunderstanding, though!
Assignee | ||
Comment 15•4 years ago
|
||
You're completely right. It only affects nightly.
Comment 16•4 years ago
|
||
(In reply to Mike Hommey [:glandium] from comment #11)
Upon reflection, we probably don't need the extra symbols anymore. Back
when we reduced the amount of stripping when enabling profiling, the
profiler wasn't able to download symbols on its own. It now is able to
do so, on all platforms.
The Gecko profiler is able to find the symbols, yes, but native profilers like Instruments and Activity Monitor are not. We use those profilers when we run into limitations with the Gecko profiler, for example for problems with high CPU usage. We often ask users who report high CPU usage on macOS to capture a sample with Activity Monitor from Nightly, and those had useful symbols in the past.
Could we undo this change for macOS?
Updated•4 years ago
|
Comment 17•4 years ago
|
||
I filed bug 1667170 about the macOS regression.
Updated•4 years ago
|
Description
•