http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstart/local-twitter/2012-12-01/2012-12-31 shows a small (< 200 ms) increase in the time to throbber start, for most phones, on Dec 22. For example, 20_64_32_4f_e9_4f_samsung_gs2_4.03 has a time of 855 at https://hg.mozilla.org/mozilla-central/rev/04abdd43606b and a time of 906 at https://hg.mozilla.org/mozilla-central/rev/ea373e534245. http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstop/local-twitter/2012-12-01/2012-12-31 shows a small (< 200 ms) increase in the time to throbber stop, for most phones, on Dec 22. For example, 78_d6_f0_cf_8d_11_nexus_s has a time of 4969 at https://hg.mozilla.org/mozilla-central/rev/04abdd43606b and a time of 5234 at https://hg.mozilla.org/mozilla-central/rev/84320dffec6e.

There is also a significant regression in the eideticker startup test at about the same time: http://eideticker.wrla.ch/#/samsung-gn/startup-abouthome-cold/timetostableframe

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=04abdd43606b&tochange=ea373e534245 throbber start and throbber stop are reported on logcat so this regression range can be narrowed manually. You could even write a script to work with hg bisect.

I'm able to reproduce this regression on my Galaxy S, so I created a bisect script. Output attached. Looks like the regression is from https://hg.mozilla.org/mozilla-central/rev/296cc3ea6e9e.

Assigning to Sriram since he was looking at a fix.

I removed the code that dealt with Tabs. Did the regression change back to the better side?

No longer an issue with bug 817706 gone.

Actually, there still appears to be a regression; see http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstop/local-twitter/2012-11-26/2013-01-25

(In reply to Geoff Brown [:gbrown] from comment #0) > http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstart/local- > twitter/2012-12-01/2012-12-31 shows a small (< 200 ms) increase in the time > to throbber start, for most phones, on Dec 22. > http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstart/local-twitter/2012-12-01/2012-12-31/notcached/noerrorbars/standarderror > For example, 20_64_32_4f_e9_4f_samsung_gs2_4.03 has a time of 855 at > https://hg.mozilla.org/mozilla-central/rev/04abdd43606b and a time of 906 at > https://hg.mozilla.org/mozilla-central/rev/ea373e534245. > > > http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstop/local-twitter/ > 2012-12-01/2012-12-31 shows a small (< 200 ms) increase in the time to > throbber stop, for most phones, on Dec 22. > http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstop/local-twitter/2012-12-01/2012-12-31/notcached/noerrorbars/standarderror > For example, 78_d6_f0_cf_8d_11_nexus_s has a time of 4969 at > https://hg.mozilla.org/mozilla-central/rev/04abdd43606b and a time of 5234 > at https://hg.mozilla.org/mozilla-central/rev/84320dffec6e. (In reply to Brian Nicholson (:bnicholson) from comment #7) > Actually, there still appears to be a regression; see > http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstop/local-twitter/ > 2012-11-26/2013-01-25 http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstop/local-twitter/2012-11-26/2013-01-25/notcached/noerrorbars/standarderror

Sorry I didn't look into this when we had builds available. There are more recent issues I think we can focus on where I can determine tight regression ranges, so I'm going to mark this as incomplete. :-(