1334762 - mozregression errors out when json-pushes request returns 404, when the cset from the bisect-target repo (legitimately) doesn't appear in an integration repo

Status: RESOLVED FIXED

(Testing :: mozregression, defect)

Description

[Ray Satiro]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:53.0) Gecko/20100101 Firefox/53.0
Build ID: 20170126004004

Steps to reproduce:

Using mozregression from git master (gui-0.9.6) to get a regression range.

setup.py install && mozregression --repo mozilla-aurora --good 20170102004004 --bad 20170126004004


Actual results:

A 404 error occurred after the offending range was found.


 5:35.94 INFO: application_buildid: 20170122071525
 5:35.94 INFO: application_changeset: 24a81d93e07cc96300f8e1f5c69034dd4743bd63
 5:35.94 INFO: application_name: Firefox
 5:35.94 INFO: application_repository: https://hg.mozilla.org/releases/mozilla-aurora
 5:35.94 INFO: application_version: 52.0a2
Was this inbound build good, bad, or broken? (type 'good', 'bad', 'skip', 'retry' or 'exit' and press Enter): good
 6:08.44 INFO: Narrowed inbound regression window from [b58eb6e9, dca7b42e] (3 revisions) to [24a81d93, dca7b42e] (2 revisions) (~1 steps left)
 6:08.44 INFO: Oh noes, no (more) inbound revisions :(
 6:08.44 INFO: Last good revision: 24a81d93e07cc96300f8e1f5c69034dd4743bd63
 6:08.44 INFO: First bad revision: dca7b42e6c67219398e6419293c075ef829adb29
 6:08.44 INFO: Pushlog:
https://hg.mozilla.org/releases/mozilla-aurora/pushloghtml?fromchange=24a81d93e07cc96300f8e1f5c69034dd4743bd63&tochange=dca7b42e6c67219398e6419293c075ef829adb29


 6:13.45 ERROR: The url u'https://hg.mozilla.org/integration/autoland/json-pushes?changeset=dca7b42e6c67219398e6419293c075ef829adb29&full=1' returned a 404 error. Please check the validity of the url.



Expected results:

I don't know

Comment 1

[Daniel Holbert [:dholbert]]

I just hit this myself. Minimal STR that trigger this for me:

  mozregression --repo mozilla-central \
    --good f985243bb630b2c78cd57731c8d8ab191aa09527 \
    --bad dd7cce059626e9787537bf9ba7863a6ad8ac06a1

This gives me the following output:
====
 0:00.08 INFO: Getting mozilla-central builds between f985243bb630b2c78cd57731c8d8ab191aa09527 and dd7cce059626e9787537bf9ba7863a6ad8ac06a1
 0:02.84 INFO: Oh noes, no (more) inbound revisions :(
 0:02.84 INFO: Last good revision: f985243bb630b2c78cd57731c8d8ab191aa09527
 0:02.84 INFO: First bad revision: dd7cce059626e9787537bf9ba7863a6ad8ac06a1
 0:02.84 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=f985243bb630b2c78cd57731c8d8ab191aa09527&tochange=dd7cce059626e9787537bf9ba7863a6ad8ac06a1

 0:03.07 ERROR: The url u'https://hg.mozilla.org/integration/fx-team/json-pushes?changeset=dd7cce059626e9787537bf9ba7863a6ad8ac06a1&full=1' returned a 404 error. Please check the validity of the url.
====

Comment 2

[Daniel Holbert [:dholbert]]

In both cases, it seems the integration repo really doesn't contain the requested changeset, for valid reasons.  So the 404 is legitimate, and mozregression should handle it gracefully.

In my case (comment 1), it looks like the "fx-team" repo stopped receiving new commits/merges in October 2016:
 https://hg.mozilla.org/integration/fx-team/pushloghtml
...so it doesn't have any new stuff since then. So that's why these changesets don't appear there.

In Ray's use-case in comment 0, the commit ID (dca7b42e6c67219398e6419293c075ef829adb29) is an Aurora-only commit (part of some merge/aurora-config stuff).  So it's correct that it doesn't appear in upstream integration repos like autoland.

wlach, could mozregression be made to handle these failures more gracefully?

Comment 3

[William Lachance (:wlach)]

(In reply to Daniel Holbert [:dholbert] from comment #2)

> wlach, could mozregression be made to handle these failures more gracefully?

Yes, it seems likely we're doing a number of silly things here (we shouldn't be supporting fx-team bisection anymore and we shouldn't look at integration branches at all when bisecting aurora). I'll look into it.

Comment 4

[Daniel Holbert [:dholbert]]

(In reply to William Lachance (:wlach) (use needinfo!) from comment #3)
> Yes, it seems likely we're doing a number of silly things here (we shouldn't
> be supporting fx-team bisection anymore

(Not for changes after Oct 2016, at least.  Seems like it could still be useful for pushlogs older than that, though, as long as we still have fx-team builds available for bisection!  Maybe it's too tricky to add that subtlety though.)

> and we shouldn't look at integration
> branches at all when bisecting aurora).

Agreed.

Thanks!

Comment 5

[William Lachance (:wlach)]

(In reply to Daniel Holbert [:dholbert] from comment #4)
> (In reply to William Lachance (:wlach) (use needinfo!) from comment #3)
> > Yes, it seems likely we're doing a number of silly things here (we shouldn't
> > be supporting fx-team bisection anymore
> 
> (Not for changes after Oct 2016, at least.  Seems like it could still be
> useful for pushlogs older than that, though, as long as we still have
> fx-team builds available for bisection!  Maybe it's too tricky to add that
> subtlety though.)

I am fairly certain that we expire inbound build data after about 4 months, so the fx-team data should be shortly gone, if it isn't already. Let's just take it out.

Comment 6

[William Lachance (:wlach)]

Attachment: PR

Hey Paul, could you have a look at this? This does two things:

1. Only bisect into inbound branches if the branch is central (bisecting to inbound from e.g. aurora doesn't make any sense)
2. Removes all reference to the obsolete fx-team branch

Comment 7

[Paul Yang [:pyang] (away)]

Comment on attachment 8836799 [details] [review]
PR

Since fx-team suspended, the patch looks good for me.

Comment 9

[William Lachance (:wlach)]

Fixed now (though note that the gui version with this fix only just made it out)