(In reply to :Felipe Gomes (needinfo me!) from comment #1)
- there are a few reasonable choices to choose for the beginning of the delta:
- process creation
- start (XPCOM start)
- main (XRE_mainInit)
Originally we had talked about using process creation, but I don't know if this will make this test also sensitive to things that we don't want to be measuring (OS-level startup stuff, for example). On the other hand, anything other than process creation will end up having red herrings due to other ordering changes (which is what this bug is meant to prevent).
I would say process creation. Or if we are concerned that the time it takes us to get the profile lock (because the previous shutdown is taking a while) would have an impact, then afterProfileLocked would be a good choice. I don't think shutdown is an issue when running Talos though.
- The sessionRestoreInit mark is only used by this test, and nothing else (at least in the codebase). Should we remove it, or is all of StartupTimeline reported to telemetry and someone might be interested in keeping this data around?
It's reported to telemetry, yes. I think we should keep it. Or removing it should be a separate discussion in its own bug. But I see value in keeping it, because the sooner we start session restore, the long the off main thread I/O has to actually read the file, and hopefully have it ready as soon as we have the first browser window displayed.