Open Bug 1372312 Opened 2 years ago Updated Last month

Make sure AWSY tests the script preloader

Categories

(Testing :: AWSY, enhancement)

Version 3
enhancement
Not set

Tracking

(Not tracked)

People

(Reporter: mccr8, Unassigned)

Details

(Whiteboard: [MemShrink:P2])

Based on some limited checking, AWSY doesn't seem to measure how much memory the script preloader uses when it is actually active.

In my local session which has been open for a bit, in the main process script-preloader is using 6MB and in the child it is using about 1.5MB.

By contrast Looking at an AWSY run, in the Start, StartSettled and TabsOpen reports, script-preloader is using about 0.6MB in the parent and 0.1MB in the content process. Furthermore, none of it is non-heap/memmapped-cache, which is where most of the memory is going in my local session.

It seems bad that we're not actually testing the memory usage of the script preloader. Does it require restarting the browser so there's something in the cache? That is what my local testing suggests is needed.
Flags: needinfo?(kmaglione+bmo)
Whiteboard: [MemShrink]
(In reply to Andrew McCreight [:mccr8] from comment #0)
> It seems bad that we're not actually testing the memory usage of the script
> preloader. Does it require restarting the browser so there's something in
> the cache? That is what my local testing suggests is needed.

In order for the mmapped cache file to be loaded, yes. But I'm not sure how valuable it would be to have that memory reported by AWSY. Aside from the header, most of it isn't currently touched after startup, so the OS is free to unmap it any time it's low on memory.
Flags: needinfo?(kmaglione+bmo)
I just want to test something that is as close as possible to the actual configuration encountered by users, in case we have some regression down the line.
Sure, that makes sense. There are a lot of other things we do differently in a fresh profile, too, so I think we should try to test on something as close as possible to a normal profile, anyway.
Seems like there's a fair amount of regressions we might miss by not testing after an initial run.
Whiteboard: [MemShrink] → [MemShrink:P1]
Component: General → AWSY
Whiteboard: [MemShrink:P1] → [MemShrink:P2]
You need to log in before you can comment on or make changes to this bug.