Git repositories for Launchpad itself

Author: William Grant
Author Date: 2024-01-23 11:55:37 UTC

Expose .buildinfo files in the API and web UI

We've been capturing these for BinaryPackageBuilds for years, and
storing them in the DB and librarian, but they haven't been available to
users.

LP: #1686242

Author: William Grant
Author Date: 2023-10-27 04:28:23 UTC

Cope more gracefully with intermittent builder glitches

buildd-manager would previously immediately count any single scan
failure against the builder and job. This meant that three glitches --
say, network timeouts -- over the course of job would result in the
build being requeued. A builder's failure count is reset on successful
dispatch, but a job's deliberately isn't since we want to fail builds
that are repeatedly killing builders. This meant that a single network
glitch in the second attempt at a build would cause it to be failed.

This added layer of failure counting substantially reduces the
likelihood of those two scenarios, by requiring five consecutive
unsuccessful scans before a single failure is counted against a builder
or job. This means that brief network interruptions, or indeed temporary
insanity on buildd-manager's part, should no longer cause builds to be
requeued or failed at all.

The only significant downside of this change is that recovery from
legitimate failures will now take a few minutes longer. But that's much
less of a concern with the very large build farm we have nowadays.

Author: William Grant
Author Date: 2023-10-27 21:59:31 UTC

Rename base failure count metric to not clash with numbercruncher

Author: William Grant
Author Date: 2023-10-27 02:13:17 UTC

Refactor buildd-manager job dispatch error handling

Previously scan sometimes caught its own errors so it could _scanFailed
with retry=False. Instead use a flag that's reset at the start of every
scan, so singleCycle can always invoke _scanFailed the right way. This
means singleCycle can always know whether a scan succeeded or failed.

Author: Colin Watson
Author Date: 2023-04-24 08:07:13 UTC

Include built assets in deployment tarballs

Merged from https://code.launchpad.net/~cjwatson/launchpad/+git/launchpad/+merge/441734

Author: William Grant
Author Date: 2020-09-16 07:57:32 UTC

Experimental _updateByHash refactoring.

Author: William Grant
Author Date: 2020-03-16 07:31:23 UTC

Wait longer in preflight.py for connections to die

Despite the comment saying it could wait for 10s, it in fact only waited
for up to 2s. This has for some time not been reliably sufficient, as
a rerun of full-update has often been necessary.

Author: William Grant
Author Date: 2020-03-05 15:08:37 UTC

Fix security.py to not crash on a role name with digits

PostgreSQL's aclitem putid emits role names unquoted in safe cases, but
security.py's regex didn't use the same safe set so failed to parse
ACLs involving usernames like "abc123".

Author: William Grant
Author Date: 2020-02-28 06:49:47 UTC

Update security.cfg for PostgreSQL 10's pgstattuple

Just eliminates security.py warnings.

Author: William Grant
Author Date: 2020-02-27 08:09:43 UTC

Index BinaryPackageReleaseDownloadCount for index-only scans

Creating an index on binarypackagereleasedownloadcount (archive,
binary_package_release, day, country, count) allows
BinaryPackagePublishingHistory.getDailyDownloadTotals,
Archive.getPackageDownloadCount and friends to execute as index-only
scans and not touch the heap at all. One sample query drops from 2789
page reads to 177.

Author: William Grant
Author Date: 2019-12-21 00:22:47 UTC

Optimise gina source and binary DB lookups

The relatively modern SourcePackagePublishingHistory.sourcepackagename
and BinaryPackagePublishingHistory.binarypackagename denormalised fields
can improve gina's plans by orders of magnitude in some cases.

This also fixes gina's binary handlers (not used on production since
2006) to correctly deal with multiple archives per distribution (it's
been incorrect since 2007's ArchiveRework).