[Buildbot-commits] [Buildbot] #2362: Timezone problem with HgPoller

[Buildbot]

#2362: Timezone problem with HgPoller
---------------------+--------------------
Reporter:  gracinet  |       Owner:
    Type:  defect    |      Status:  new
Priority:  major     |   Milestone:  0.8.7
 Version:  master    |  Resolution:
Keywords:            |
---------------------+--------------------

Comment (by gracinet):

 Okay, my bad, a typical case of double correction:

 {{{
 >>> time.gmtime(1346158664)
 time.struct_time(tm_year=2012, tm_mon=8, tm_mday=28, tm_hour=12,
 tm_min=57, tm_sec=44, tm_wday=1, tm_yday=241, tm_isdst=0)
 >>> time.localtime(1346158664)
 time.struct_time(tm_year=2012, tm_mon=8, tm_mday=28, tm_hour=14,
 tm_min=57, tm_sec=44, tm_wday=1, tm_yday=241, tm_isdst=1)
 }}}

 So, the first number of hg log is the UTC timestamp (and the general
 convention is that timestamps are always UTC, since the epoch is expressed
 in UTC).
 The second number is an indication of the original timezone, of which
 buildbot does not care. By summing the two, !HgPoller is plain wrong.

 Prior checkings were done with ``datetime.fromtimestamp`` which actually
 converts to local time, but as a naive datetime instance, whence the
 confusion. Should have read more carefully the doc, sorry.

 I will propose a fix ASAP.

-- 
Ticket URL: <http://trac.buildbot.net/ticket/2362#comment:3>
Buildbot <http://buildbot.net/>
Buildbot: build/test automation