1214347 - Figure out how to pass MAR URLs to l10n partial generator

Status: RESOLVED FIXED

(Release Engineering :: Release Automation, defect)

Description

[Rail Aliiev [:rail]]

When we generate a release promotion graph, we don't yet know the final URLs of complete MARs generated by desktop_l10n.py.

We have the following options to define these URLs:

1) Use FTP, err, S3 URLs based on pre-defined URL pattern, something like 
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2015/10/2015-10-11-06-35-47-date-l10n/firefox-44.0.gu-IN.linux-i686.complete.mar

pros: relatively easy to implement, doesn't require AUTH, the URL is predefined
cons: hard coded URLs are bad

2) Use TC Index to calculate the corresponding task IDs and then use them to retrieve the corresponding artifacts:
Something like:
 * GET https://index.taskcluster.net/v1/task/gecko.v2.date.revision.a9320053068dd168603884e5e3b1062d193fd048.firefox-l10n.linux-opt.he
 * use taskId to generate the URL *run time* https://queue.taskcluster.net/v1/task/CqawVj44RW2GtM5_icXgLw/runs/0/artifacts/public/build/firefox-44.0.he.linux-i686.complete.mar

Pros: relatively easy to implement, uses TC artifacts
Cons: requires buildbot2treeherder platform conversion (we can add another route to fix this)

Comment 1

[Rail Aliiev [:rail]]

Done Done Done