If a PR is *merged*, enqueue it for deletion (with a 2 weeks delay).
Mainly to avoid FW branches staying around long after they've been
merged (possibly eventually closed?), will also clean up regular
merged branches, including historical merges forgotten by their
author.
Fixes#230
When a build only create sub-builds, the build_time is verry small (a few seconds),
and this information is not relevant. This commit propagates end_time to parent_build
if parent_build is done or running.
In the case where we have a co-dependent forward port (co-dependent
PRs got forward-ported, which they should be together) where *one* of
the PRs got explicitly updated, the batch would "fall into a hole"
being handled as neither "this is part of a forward-port sequence" nor
"this is a new merge to forward-port" (the latter being the proper
one).
Modify & remove guards which checked that either no or all PRs in a
batch have parents: should be either all or not all.
Fixes#231
Turns out tagging PRs requires a pretty significant level of ACLs
which we may not want to give to the forwardbot?
Anyway use the mergebot ACLs (which already include tagging) for this.
* add a sorted method on fake models
* fix recordset equality to ignore ids order
* when creating commits on a ref, add a param to only *update* the ref
(forcefully): when simulating a force-push we don't want to *create*
a ref as that might silently be done in the wrong repository entirely
* fix pytest.skip call at the module level, not sure where it came
from and why I missed it until now
The closing or reopening of PRs was not logged at all, which can be
inconvenient when trying to find out why PRs are closed (or not) in
the backend.
Also leverage PR display_name improvements from
3ce3dd9569 for more regular PR names in
logs.
Due to the title formatting of FP PRs, we'd get incorrectly formatted
commit messages if the PR was *merged* (rather than squashed /
fast-forwarded) due to either "merge" or "rebase-merge" integration
mode: in that case the PR message would be used as message for the
merge commit and that'd be along the lines of "Forward Port of #xxx to
<somebranch> (failed)", followed by the old PR message (e.g. see this
commit message).
* re-extract and reuse original PR title, just prefix with "[FW]"
* finally add support for tagging, and use that to tag the PRs,
especially for the failed / conflict marker which is quite important
Closes#229
When posting a reminder that there are open / waiting forward ports on
a source PR, also post *which* PRs those are.
While at it, move the cron code in a proper python file (so we can use
stuff from odoo.tools), and fix display_name so we can straight use
display_name as a github ref' ({owner}/{repo}#{number}). This impacts
log-grepping but it seems like an improvement nonetheless.
Closesodoo/runbot#228
* shorten the postfix, forwardbot is now a bigram!
* shorten the uniquifier: go from 5 to 3 bytes, and use urlsafe base64
that way we only have a 4-char uniquifier instead of 8
* while at it, fix deprecated calls to logging.warn (should be
logging.warning)
Fixes#226
The fw-bot testing API should improve the perfs of mergebot tests
somewhat (less waiting around for instance).
The code has been updated to the bare minimum (context-managing repos,
change to PRs and replacing rolenames by explicit token provisions)
but extra facilities were used to avoid changing *everything*
e.g. make_commit (singular), automatic generation of PR refs, ...
The tests should eventually be updated to remove these.
Also remove the local fake / mock. Being so much faster is a huge
draw, but I don't really want to spend more time updating it,
especially when fwbot doesn't get to take advantage. A local /
lightweight fake github (as an external service over http) might
eventually be a good idea though, and more applicable (including to
third-parties).
The queue would get items to process one at a time, process, commit,
and go to the next. However this is an issue if one of the item fails
systematically for some reason (aka it's not just a transient
failure): the cron fails, then restarts at the exact same point, and
fails again with the same issue, leading to following items never
getting processed.
Fix by getting all the queue contents at once, processing them one by
one and "skipping" any item which fails (leaving it in place so it can
get re-processed later).
That way, even if an item causes issues, the rest of the queue gets
processed normally. The interruption was an issue following
odoo/enterprise#5670 not getting properly updated in the
backend (backend didn't get notified of the last two updates /
force-push to the PR, so it was trying to forward-port a commit which
didn't exist - and failing).
Attempt to avoid some of the comment spam by dedup-ing input (only
signaling when the status actually changes and ignoring identity
transformations) and in case of failing CI keeping the last failed
status and not signaling on the next update if it's the same failure.
Closes#225
* only show 2 stagings on cellphones as 4 is way too much, moving to a
vertical layout would probably be a bad idea as stagings can already
be very tall and then we have multiple branches stacked on one
another, unless we also make branches foldable
the more complete list of stagings (per branch) is available on the
branch's page anyway so providing a not-completely-broken home looks
more useful, and at a fundamental level the current / last staging
is really the one we care about
* remove the size bounds on stagings to avoid smushing all the cells
together and overlapping text, sadly can't overflow scroll the
stagings element because you can't have an overflow-x: scroll and an
overflow-y: visible (that becomes auto)
When running tests on some machines, it's apparently possible for the
PR-creation webhook to come back before the PR creation request has,
leading to the creation of the PR from the API call duplicating that of
the webhook and blowing up.
To fix, immediately commit the transaction then check if we already have
the PR we just created in the system, and only create it explicitly if
not.
User should probably be warned when they try to set the limit ("up
to") in a context where it's going to be ignored:
* on a forward port PR (should be set on the source)
* on a merged source (should be set before the PR is merged)
Closes#213
The staging validation routine would ignore stagings which were
cancelled or ff_failed, but it should also have ignored failed and
successful aka all terminal state.
Simplify the condition for that: just ignore a staging's validation if
the staging is not pending.
Closes#211
In selecting the parent commits to list on the last PR, we would miss
the *first* forward-port of the sequence. Not sure why we added a
detrimental check on source_id there.
Also add a missing space between "chain" and "containing" in the case
where there's at least one forward-port PR other than the final one.
Fixes#212
Test is probably more complex than necessary (thinking about it, the
failed staging is probably unnecessary) but that triggers the issue
and matches the original scenario.
The problem was really with new CI events being received on the last
forward-port PR of a sequence: previous PRs would have a child PR so
the check would abort, but for the last PR it would go through, fail
to find an active batch, then blow up as it tries to create a
forwardport.batch without an actual batch.
Change this to use the existence of an inactive batch not linked to a
staging as a flag that the PR has been processed and forward-ported.
Closes#218
Turns out we don't want to close the cursor on success, we just want to
commit, but that's not what the default context manager does.
So don't use said context manager.
If a _validate call blows up, the entire Commit._notify cron gets
stuck, which is an issue because not only does it stop creating
forward ports, it also stops "progressing" stagings.
If the CI is greatly backed up (either insufficient capacity or jobs
spike) a timeout which is normally perfectly fine might be
insufficient e.g. given a 2h timeout, if a job normally takes 80mn but
the staging's job starts 40mn after the staging was actually created
we're sunk. And cancelling the staging once the job has finally gotten
started is not going to improve load on the CI, it just wastes a CI
slot.
Therefore assume a `pending` event denotes the actual start of the job
on the CI, and reset the timeout to start from that moment so
ci_timeout is the timeout of the CI job itself, not of the staging
having been created.
Closes#202
Converge the pytest setups of runbot_merge and forwardport a bit
more (the goal is obviously to eventually share the infrastructure so
they run the same way).
If a PR is explicitly updated, it gets converted to a normal
PR[0]. Before this, users had no indication that this had happened and
might be wondering what they're supposed to do (or try to r+ via the
forwardbot, which doesn't work on a root PR).
[0] to an extent: the PR still has a source and might have children,
in which case the followups will be created from the source &
existing followups should be updated to match
Closes#206
When creating the conflicting commit of a single commit PR, reuse the
original commit's meta-information so the developer / fixer can more
easily update it in-place.
Closes#204
In the case where an FP sequence is interrupted (e.g. there was a
conflict during one of the intermediate steps), followups get linked
to the original source but don't get linked to the "interruption" PR
which is a bit confusing.
Link FP PRs to both source and root if they're different.
* always allow specifying the PR's own branch as a forward-port limit
/ target (even if deactivated or disabled)
* add an "ignore" alias to "up to <pr target>" for clarity
* add dedicated feedback when deactivating forward-port of a PR
Fixes#191
Having all the feedback be sent by the mergebot user (github_token) is
confusing. Add a way to specify which field of project should be used to
source the token used when sending feedback.
Fixes#190
When a build_error active field is changed, the onchange leads to a
traceback. Anyway, the onchange was not a good idea as it only reflects
UI changes.
With this commit, the write method is overwritten to change the
child_ids active fields too. Also, the active_test context is used to
correctly compute the childs_ids and children_build_ids.
A test is also added for all that.
The new feature in odoo/enterprise#4879 needs the firebase-admin
package. As it cannot be added to the requirements.txt, the package is
added in the Dockerfile to be able to test it on the runbot instances.
- Add a keep running flag on the build to allow a build to stay in
running state until the flag is switched off ( or the build killed)
- Do not update configs and config_steps data
- Add a first/last_seen_build and first/last_seen_date on build.error
- Children error builds now include the parent builds too
- Use a notebook on build.error form view to display builds and linked
errors
- Update result when a build triggers a change from 'warn' to 'ko' too
- Add the sticky flag on the error logs stored sql view
When a build error appears with the same fingerprint as already known
one which was supposedly fixed, the build is simply added to the known
build error.
In order to keep an eye on such reappearing bugs and keep the fixing
history separated, this commit simply creates a new build_error.
Old build errors with the same hash (or child_ids 's hashes) appears in
a computed field error_history_ids.
When using the runbot frontend, it's sometimes very frustrating when
trying to copy branch name, some mouse gym is necessary.
With this commit, a copy to clipboard button is added near the branch
name on the frontend.
* don't warn on every PR on the dot every week, instead wait for the
PRs to be "sufficiently old" (at least 3 days)
* after discussion with bugfix, the reminder ping should be sent every
day following the PR being "old enough"
* run the cron every day instead of every week
* add an override context key for test purposes
Closes#198
Working copies were created in tempdir under the assumption that
they're, well, temporary.
However after thinking about it more there are two issues with this:
* tempdirs might not be in the same FS as the cache dir, meaning
meaning `git clone` can't hardlink the repo objects and has to
copy them
* tempdirs are often on RAM-backed tmpfs, which is not great when we're
filling them with multiple GB worth of git repository...
If the default limit of a forward-port sequence is not a valid
target (either disabled or not actually forward-ported to), the last
effective forward port in a sequence will be commented on as any
intermediate PR rather than get a proper ping and r+ instructions.
Also remove a bunch of leftover prints in the tests.
Fixes#192
