1390239 - Artifact build of df9beb7 wants to download installers from a 3 day old build

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[Henrik Skupin [:whimboo][⌚️UTC+1]]

I noticed this today because log spam from bug 1388748 was still present for me in my local build on OS X for mozilla-central. From the log it looks like instead of using the current changeset, a 3 day old one was used:

> $ hg up -C central
> 0 files updated, 0 files merged, 0 files removed, 0 files unresolved
> (df9beb7)henrik@whimboo-mac:/data/code/gecko%

So I'm on https://hg.mozilla.org/mozilla-central/rev/df9beb7.

Running `mach build` results in:

> 0:04.61 Attempting to find a pushhead containing 58f678547059f27e2dd930b7c9fa4b7bf44365e2 on mozilla-central.
> 0:04.61 Attempting to find a pushhead containing 58f678547059f27e2dd930b7c9fa4b7bf44365e2 on integration/mozilla-inbound.
> 0:04.62 Attempting to find a pushhead containing 58f678547059f27e2dd930b7c9fa4b7bf44365e2 on releases/mozilla-beta.
> 0:05.70 Retrieving the last 50 pushheads starting with id 93626 on integration/mozilla-inbound
> 0:05.70 Retrieving the last 50 pushheads starting with id 32318 on mozilla-central
> 0:05.75 Installing from remote pushhead 58f678547059f27e2dd930b7c9fa4b7bf44365e2 on integration/mozilla-inbound

This is https://hg.mozilla.org/mozilla-central/rev/58f678547059f27e2dd930b7c9fa4b7bf44365e2 which landed Friday Aug 11th on mozilla-central.

Why was the changeset on mozilla-central not found, and instead inbound used? And then a 3 day old build?

Chris currently have a look into this locally. Thanks.

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+1]]

With f31c029 checked out locally I get:

> $ hg log --template '{node}\n' -r 'last(public() and ::., 5)'
> 1b9dac0c4cf374c8fc26bcb432302dfd293098c7
> 31b5a863aee278ac3c39db8d142a7fa98d38797e
> ffc2e43aa834e05f0d51d68dfb36317c1b408b08
> 8be5e7c145999f436c84b181af44838cd28a9efb
> f31c029bbd16cd779f191a63065260aaafad026b

I upgraded Mercurial by end of last week to version:

% hg --version
Mercurial Distributed SCM (version 4.3.1)

Comment 2

[Chris Manchester (limited bugmail, email directly)]

I can reproduce this locally, it looks like our mercurial revset to try to find the most recent public changeset isn't giving us what we want. The revset in question is: http://searchfox.org/mozilla-central/rev/6482c8a5fa5c7446e82ef187d1a1faff49e3379e/python/mozbuild/mozbuild/artifacts.py#980

It seems possible there was a change in the mercurial update we need to take into account, I'll take a look.

Comment 3

[Chris Manchester (limited bugmail, email directly)]

Attachment: Bug 1390239 - Accommodate any ordering of "last" revset in artifact builds.

A recent mercurial update changed the order of revisions returned by the
"last" revset. The expected revisions are still output, so the artifact
code is updated in this change to impose its own order based on the local
revision order to accommodate any output order.

Review commit: https://reviewboard.mozilla.org/r/168400/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/168400/

Comment 4

[Gregory Szorc [:gps]]

Comment on attachment 8897092 [details]
Bug 1390239 - Accommodate any ordering of "last" revset in artifact builds.

https://reviewboard.mozilla.org/r/168400/#review173638

::: python/mozbuild/mozbuild/artifacts.py:987
(Diff revision 1)
>              self._hg, 'log',
> -            '--template', '{node}\n',
> +            '--template', '{rev}:{node}\n',
>              '-r', 'last(public() and ::., {num})'.format(
>                  num=NUM_REVISIONS_TO_QUERY)
>          ], cwd=self._topsrcdir).splitlines()
> +        return [i.split(':')[-1] for i in reversed(sorted(last_revs))]

Nit: `sorted(last_revs, reverse=True)`.

Comment 5

[Chris Manchester (limited bugmail, email directly)]

Comment on attachment 8897092 [details]
Bug 1390239 - Accommodate any ordering of "last" revset in artifact builds.

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/168400/diff/1-2/

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/67caba0c3909
Accommodate any ordering of "last" revset in artifact builds. r=gps

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/67caba0c3909