1244143 - Record whether or not an artifact build was used in build telemetry data

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[Dan Minor [:dminor]]

We need to record whether or not an artifact build was used in the telemetry data - artifact builds should be faster, and we'd like to know how many people are using them.

Comment 1

[Nick Alexander :nalexander [he/him]]

dminor: thanks for filing this.  How much of the local mozconfig are we comfortable capturing?  For @mozilla.com addresses, I hope it's a lot.  If we did this, it would let us know if --enable-artifact-builds was enabled; but it might also be helpful to know how many folks put their objdir in the srcdir, for example (which has pros and cons); or how many people are using dated options; or ...

Comment 2

[Dan Minor [:dminor]]

That's a good point. My first thought is that we are better off making a list of parsed values we'd like to extract from mach rather than just slurping up the whole file and risk accidentally getting something we shouldn't see, but maybe I'm being too conservative here.

Comment 3

[Chris Manchester (limited bugmail, email directly)]

Hopefully we could submit the entire build config (substs, defines, topscrdir, topobjdir, etc.). Inputs to the build all seem within scope.

Comment 4

[Dan Minor [:dminor]]

(In reply to Chris Manchester [:chmanchester] from comment #3)
> Hopefully we could submit the entire build config (substs, defines,
> topscrdir, topobjdir, etc.). Inputs to the build all seem within scope.

That makes sense to me. I'll morph the bug accordingly.

Comment 5

[Dan Minor [:dminor]]

Attachment: MozReview Request: Bug 1244143 - Record whether or not an artifact build was used in build telemetry data r?gps

This adds the defines, substs, topoutdir and topsrcdir configuration values
to the telemetry data we are currently collecting. This adds about 70k of
data, but probably ends up being better than having to iterate on adding
values we care about piecemeal.

Review commit: https://reviewboard.mozilla.org/r/33463/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/33463/

Comment 6

[Gregory Szorc [:gps]]

Comment on attachment 8715460 [details]
MozReview Request: Bug 1244143 - Record whether or not an artifact build was used in build telemetry data r?gps

https://reviewboard.mozilla.org/r/33463/#review30173

::: python/mozbuild/mozbuild/mach_commands.py:501
(Diff revision 1)
> +            telemetry_data['defines'] = self.defines
> +            telemetry_data['substs'] = self.substs
> +            telemetry_data['topsrcdir'] = self.topsrcdir
> +            telemetry_data['topobjdir'] = self.topobjdir

The build config data may contain PII, notably filesystem paths, which may contain a username. 

People could also argue that this data could be used to gleen information about a user's machine to help construct an attack. e.g. it may identify they are running a known vulnerable version of some piece of software. This data makes the Telemetry database an attactive target for an attacker wishing to compromise a Firefox developer's machine.

As much as I want to record all of the things, I think we should cherry pick values for now. I don't think mass capturing of the full build configuration is warranted. Let's drop this for now.

Comment 7

[Dan Minor [:dminor]]

Fair enough, I'll morph this back to the original bug, which is to capture whether an artifact build was done, which should be safe.

Comment 8

[Dan Minor [:dminor]]

Comment on attachment 8715460 [details]
MozReview Request: Bug 1244143 - Record whether or not an artifact build was used in build telemetry data r?gps

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/33463/diff/1-2/

Comment 9

[Gregory Szorc [:gps]]

Comment on attachment 8715460 [details]
MozReview Request: Bug 1244143 - Record whether or not an artifact build was used in build telemetry data r?gps

https://reviewboard.mozilla.org/r/33463/#review30755

Comment 10

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/b06f5fef9488

Comment 11

[Carsten Book [:Tomcat]]

sorry had to back this out , since this caused trouble on 10.7 opt like https://treeherder.mozilla.org/logviewer.html#?job_id=21427471&repo=mozilla-inbound

Comment 12

[Pulsebot]

Backout:
https://hg.mozilla.org/integration/mozilla-inbound/rev/2064cff635eb

Comment 13

[Dan Minor [:dminor]]

Comment on attachment 8715460 [details]
MozReview Request: Bug 1244143 - Record whether or not an artifact build was used in build telemetry data r?gps

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/33463/diff/2-3/

Comment 14

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/4137fa2415c8

Comment 15

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/4137fa2415c8