This was bug 1006478, no? - Kyle
On Tue, Jun 10, 2014 at 8:20 AM, Vladan Djeric <[email protected]> wrote: > User Advocacy reports that users of Firefox 29 are seeing Firefox take a > long time to shut down and that they are getting the "Firefox is already > running" error message (i.e. a locked profile) when they try to relaunch > Firefox. > > Can anyone think of any bugs or patches from that time frame that might be > responsible? > > --- > > We analyzed some of the Telemetry data to try to identify the source of this > regression: > > 1. The time it takes Firefox to shut down "cleanly" (i.e. successfully) is > basically unchanged across versions: http://mzl.la/1hCbMMX > > 2. The distribution of clean shutdown times in Firefox 27+28+29 has an odd > secondary peak at 5-8 seconds: > > * Firefox 26 has a normal distribution: http://mzl.la/1hCezFO (see the > histogram on the bottom of the page) > * Firefox 27 has a secondary peak between 5 & 8 seconds: > http://mzl.la/1hCeFxb > * Firefox 28: http://mzl.la/1hCeHVO > * Firefox 29: http://mzl.la/1qkiXcD > > 3. We have information on the number of startups that failed because of a > locked profile via the SIMPLE_MEASURES_FAILEDPROFILELOCKCOUNT Telemetry > probe. These failed startups are preceded by an unclean shutdown. > > There was an an increase in the proportion of "failed startups" from Firefox > 28 to Firefox 29 although there was also a major improvement in Firefox 26: > > * Firefox 24: 2.46M startups were preceded by at least one "locked profile" > failed startup out of a total 319.58M normal startups. This is a failure > rate of 0.770% > * Firefox 25: 3.16M / 406.83M = 0.777% > * Firefox 26: 3.33M / 468.88M = 0.710% > * Firefox 27: 1.26M / 422.52M = 0.298% > * Firefox 28: 1.03M / 341.73M = 0.301% > * Firefox 29: 276.08K / 71.56M = 0.385% > > In other words, the proportion of failed startups increased by a quarter in > Firefox 29 vs Firefox 28. > > 4. We did not find any correlations between users getting the "locked > profile" message and any particular extension. There is also no correlation > with any Flash plug-in version. > > 5. We also looked at the stacks of temporary hangs (aka "chrome hangs") > occurring around shutdown time and we did not find any smoking guns there. > > Any ideas? > _______________________________________________ > dev-platform mailing list > [email protected] > https://lists.mozilla.org/listinfo/dev-platform _______________________________________________ dev-platform mailing list [email protected] https://lists.mozilla.org/listinfo/dev-platform
