Florian Quèze
Florian Quèze
The stack is technically correct, but it's making it hard to compare stacks because as soon as something spins the event loop, any event can occur and be attached to...
"Focus on function" is something we also looked at but it excludes the other samples. It would be nice to have the same transform as "Focus on function" for samples...
Unpacking the discussion even more: I was saying that to be able to compare profiles from Talos, I would need to simplify the stacks so that stacks that are identical...
I think that would be useful, yes. Of course without having it it's hard to say how often I would actually use it. ProcessNextEvent was one example of use case....
Just finding a place to display the "50ms" value so that the red jank bars feel less like magic for people discovering the profiler would already be an improvement.
With the new responsiveness tracks, it might be easier to access this information now.
> can I give it a shot? @fqueze Sure, go ahead :). (@julienw marked the issue as good first issue so I expect him to reply; sorry for the delay)
Is this specific to android? I could be tempted to merge a samply profile with a resource usage profile.
> You are probably interested in keeping markers from both profiles, which will be more complicated because you'll also have to merge the schemas. If we added more useful markers...
I don't know how the numbers stored in browsertime.json are computed. It's likely a subset of the profile, based on a start time and stop time for an interesting range...