Closed Bug 1431640 Opened 3 years ago Closed 3 years ago

Unable to apply updates, always failing with .mar size mismatch

Categories

(Thunderbird :: General, defect)

defect
Not set
major

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: gerard-majax, Unassigned)

Details

Attachments

(4 files)

Attached file JS console output
This has been a recurring issue for a few days for me.

STR:
 1. Open About dialog,
 2. Check for update,

Expected:
 Download update and apply it properly

Actual:
 Complains that the update cannot be applied.

Console shows some constant 24-bytes mismatch when checking the size of the file. Manually downloading the .tar.bz2 updates properly (this is how I updated twice in the past week).
Attached file served update.xml
I've downloaded https://archive.mozilla.org/pub/thunderbird/nightly/2018/01/2018-01-18-06-54-19-comm-central-l10n/thunderbird-59.0a1.fr.linux-x86_64.complete.mar from multiple places, they all have the same size (46213803) and sha512 (0f18e27518caa96e1b83044c05c4b788018a6189f5c26bf5cae51f39739b4a7e245bb744db1b34e818747c24894538cbc870773a0943088f76b158943a55cea5). So it would seems like the XML exposed by AUS is bogus?
Flags: needinfo?(mozilla)
Severity: normal → major
It turns out that in Bug 1423313, the builds that were supposed to generate L10N builds for 32-bit linux were actually generating builds for 64-bit linux and uploading files with the 64bit file name. This, since 2017-12-05, the nominally 32bit builds and the real 64bit builds have been racing to upload files, with last one wins. When the 32bit build won, the update information for 64 bit builds would point to the overwritten file.
Flags: needinfo?(mozilla)
Pushed by mozilla@hocat.ca:
https://hg.mozilla.org/comm-central/rev/9dd7d3ad611c
Actually repack 32bit Linux binaries on linux32 L10N builders; r=me CLOSED TREE
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
Pushed by mozilla@hocat.ca:
https://hg.mozilla.org/comm-central/rev/c5dc71e950f1
Repack win64 builds when requested to do so; r=me
You need to log in before you can comment on or make changes to this bug.