Re: status of performance regressions on mozilla-central

L. David Baron writes:

>  * sdwilsh's backout of 456910 and the sqlite changes fixed the
>    performance regression on Mac OS X 10.4 and 10.5 in the one build
>    generated right after he landed the backout, and then in the next
>    cycle (which only added a trivial reftest.list change), the
>    performance numbers were *even better*, more than erasing the
>    increases from yesterday.
>
>  * on Windows XP, the builds following the landing of the places
>    fsync work got us back the 20ms that we lost yesterday, but when
>    that was backed out the needle didn't move.  Then when sdwilsh
>    backed out the sqlite upgrade and 456910, we got 20ms back again,
>    so again, we're a good bit ahead of where we were yesterday
>    morning.  In other words, we're also ahead of where we started
>    yesterday morning, but the steps happened at different times.

Being "even better" and "a good bit ahead" I think are due to
changes in the testing infrastructure:

https://bugzilla.mozilla.org/show_bug.cgi?id=457582

(I haven't looked into Vista/Txul, but I suspect that is another issue.)