Closed
Bug 1194937
Opened 10 years ago
Closed 10 years ago
mozregression default 'good' date 2009-01-01 doesn't work on win64
Categories
(Testing :: mozregression, defect)
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: raysatiro, Assigned: parkouss)
Details
Attachments
(1 file)
User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:43.0) Gecko/20100101 Firefox/43.0
Build ID: 20150811011940
Steps to reproduce:
Run mozregression with no parameters
Actual results:
It always returns '503 Server Error: Server Too Busy'
0:00.59 LOG: MainThread INFO bits option not specified, using 64-bit builds.
0:00.59 LOG: MainThread INFO No 'bad' date specified, using 2015-08-14
0:00.59 LOG: MainThread INFO No 'good' date specified, using 2009-01-01
0:07.47 LOG: MainThread Build Finder WARNING Got HTTPError - retrying
0:07.47 LOG: MainThread Build Finder WARNING 503 Server Error: Server Too Busy
0:07.47 LOG: MainThread Build Finder WARNING Got HTTPError - retrying
0:07.47 LOG: MainThread Build Finder WARNING 503 Server Error: Server Too Busy
[...]
1:44.64 LOG: MainThread Bisector INFO Downloading build from: https://archive.m
ozilla.org/pub/mozilla.org/firefox/nightly/2013/01/2013-01-04-03-08-23-mozilla-c
entral/firefox-20.0a1.en-US.win64-x86_64.zip
EOF occurred in violation of protocol (_ssl.c:590)
1:44.85 LOG: Dummy-1739 Bisector INFO Last good revision: c458465578e0
1:44.85 LOG: Dummy-1739 Bisector INFO First bad revision: 7649ffe28b67aa2dad0f6
7ea01500c0ff91b2bac
1:44.85 LOG: Dummy-1739 Bisector INFO Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=c458465578e0&tocha
nge=7649ffe28b67aa2dad0f67ea01500c0ff91b2bac
1:44.85 LOG: Dummy-1739 Bisector INFO To resume, run:
1:44.85 LOG: Dummy-1739 Bisector INFO mozregression --good=2010-05-28 --bad=201
5-08-13 --app=firefox --bits=64
Another thing that happens is it may just quit without any of the info above:
1:02.53 LOG: MainThread Build Finder WARNING 503 Server Error: Server Too Busy
Retrieving valid builds from datetime.date(2009, 12, 7) generated an exception:
EOF occurred in violation of protocol (_ssl.c:590)
Expected results:
I think it should default to a date that works like 2010-05-28
Assignee | ||
Comment 1•10 years ago
|
||
Hmm, it works for me, at least it find builds on Linux 64:
mozregression
/home/jp/.mozregression.cfg loaded
0:00.39 LOG: MainThread INFO No 'bad' date specified, using 2015-08-15
0:00.39 LOG: MainThread INFO No 'good' date specified, using 2009-01-01
0:07.69 LOG: MainThread Bisector INFO Downloading build from: https://archive.mozilla.org/pub/mozilla.org/firefox/nightly/2012/04/2012-04-23-03-07-00-mozilla-central/firefox-14.0a1.en-US.linux-x86_64.tar.bz2
^C=== Downloaded 7% =====
Interrupted.
I'll try on windows, maybe there are no builds and the mozregression stress the server for nothing.
Also, "Server too busy" may indicate that the server was not in a good state at the time you tried.
Comment 2•10 years ago
|
||
I see this too, FWIW.
Comment 3•10 years ago
|
||
Looks like it finally got through and downloaded a 2013-01-08 build.
Assignee | ||
Comment 4•10 years ago
|
||
Ok, I suspect this is because you are both using a win 64 platform, and win64 builds were not present at this time.
Do you have better results if you use the --bits=32 flag ?
If yes, I think the best solution would be for mozregression to default to the date of the first 64 win build (if using that platform). I will search that date.
Assignee | ||
Comment 5•10 years ago
|
||
Ok, so I found a build on 2010-05-28 (https://archive.mozilla.org/pub/mozilla.org/firefox/nightly/2010/05/2010-05-28-06-mozilla-central/). I think it is the oldest win 64 m-c nightly we have.
If I use that date with mozregression on my linux box, e.g:
mozregression -g 2010-05-28
> 0:00.36 LOG: MainThread INFO No 'bad' date specified, using 2015-08-23
> 0:06.81 LOG: MainThread download INFO Downloading build from: https://archive.mozilla.org/pub/mozilla.org/firefox/nightly/2013/01/2013-01-09-03-09-42-mozilla-central/firefox-21.0a1.en-US.linux-x86_64.tar.bz2
> ^C=== Downloaded 0% =====
>
> Interrupted.
> 0:08.60 LOG: Dummy-13 Bisector INFO Last good revision: aa20e4f73d81
> 0:08.60 LOG: Dummy-13 Bisector INFO First bad revision: 22c34579ae0720e7d3dc39a22b9d33f13bc0198b
> 0:08.60 LOG: Dummy-13 Bisector INFO Pushlog:
> https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=aa20e4f73d81&tochange=22c34579ae0720e7d3dc39a22b9d33f13bc0198b
>
> 0:08.60 LOG: Dummy-13 main INFO To resume, run:
> 0:08.60 LOG: Dummy-13 main INFO mozregression --good=2010-05-28 --bad=2015-08-23 --app=firefox --bits=64
It will download a build for 2013-01-09. It seems quite good if I compare with comment 3 (the one day difference is probably due to some nightly build file missing in either win64 or linux64).
So I suspect this is the root cause. and using "2010-05-28" as a default good date for win 64 should resolve the issue.
Assignee | ||
Comment 6•10 years ago
|
||
(In reply to Ray Satiro from comment #0)
> User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:43.0) Gecko/20100101
> Firefox/43.0
> Build ID: 20150811011940
...
> Expected results:
>
> I think it should default to a date that works like 2010-05-28
Oh! I missed that in the first comment (not sure how)! Thanks Ray Satiro for pointing this. :)
So I totally agree, this is the date we should use for win64 bisection.
Assignee: nobody → j.parkouss
Status: UNCONFIRMED → ASSIGNED
Ever confirmed: true
Assignee | ||
Updated•10 years ago
|
Summary: mozregression default 'good' date 2009-01-01 doesn't work → mozregression default 'good' date 2009-01-01 doesn't work on win64
Assignee | ||
Comment 7•10 years ago
|
||
Attachment #8651517 -
Flags: review?(wlachance)
Comment 8•10 years ago
|
||
Comment on attachment 8651517 [details] [review]
mozregression default 'good' date 2009-01-01 doesn't work on win64
It would be nice if there were some way of fetching reasonable dates for these things from the server as opposed to hardcoding them in the client, but this looks good as a stopgap measure.
Attachment #8651517 -
Flags: review?(wlachance) → review+
Assignee | ||
Comment 9•10 years ago
|
||
(In reply to William Lachance (:wlach) from comment #8)
> It would be nice if there were some way of fetching reasonable dates for
> these things from the server as opposed to hardcoding them in the client,
> but this looks good as a stopgap measure.
Hmm, yeah. Maybe something for taskcluster ? or add a json file somewhere on ftp with such information ?
Well for now, I landed it in https://github.com/mozilla/mozregression/commit/570a9fb293295a1520b5c578e4b860e99f186f55.
Status: ASSIGNED → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•