BCI repo publisher: also trigger a rebuild of devel:BCI:$VERSION #3023
Conversation
| from osc.core import makeurl | ||
| from osc.connection import http_GET | ||
|
|
||
| _SLE_VERSION_T = Literal['15-SP3', '15-SP4', '15-SP5', '15-SP6'] |
There was a problem hiding this comment.
Does this really need to be typed?
There was a problem hiding this comment.
No it doesn't have to, but I found it quite useful, because I was wondering what the values of this variable are. Initially I thought this was a package version only to realize later that it's actually 15-SP$N.
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #3023 +/- ##
=======================================
Coverage 28.46% 28.46%
=======================================
Files 85 85
Lines 14722 14722
=======================================
Hits 4191 4191
Misses 10531 10531 ☔ View full report in Codecov by Sentry. |
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
34cfa12
to
6adbac2
Compare
gocd/bci_repo_publish.py
Outdated
| req.raise_for_status() | ||
|
|
||
| for pkg in self.fetch_package_names_in_project(build_prj): | ||
| url = makeurl(self.apiurl, ['trigger', 'rebuild'], {'project': build_prj, 'package': pkg}) |
There was a problem hiding this comment.
Ugh, no, please no fanout of api posts against the build service. we don't need DoS here. "'package" can take a list, so if project wide rebuild is not working (which it should) then please do one call with the "package" being a list of packages to rebuild.
There was a problem hiding this comment.
Sadly, a list doesn't work.
❯ curl -H "Authorization: Token $TOKEN" -X POST "https://api.opensuse.org/trigger/rebuild?project=devel:BCI:SLE-15-SP5&package=rmt-helm,sle15-kernel-module-devel,rust-oldstable-image,rust-stable-image"
<status code="unknown_package">
<summary>Package not found: devel:BCI:SLE-15-SP5/rmt-helm,sle15-kernel-module-devel,rust-oldstable-image,rust-stable-image</summary>
</status>And specifying the parameter multiple times rebuilds just one package:
❯ curl -H "Authorization: Token $TOKEN" -X POST "https://api.opensuse.org/trigger/rebuild?project=devel:BCI:SLE-15-SP5&package=rmt-helm&package=sle15-kernel-module-devel&package=rust-oldstable-image&package=rust-stable-image"
<status code="ok">
<summary>Ok</summary>
</status>So sadly, we need to DoS OBS 🤷
There was a problem hiding this comment.
Dirk is most likely confusing it with the route used by rebuildpac. The rebuild token is a very special route
There was a problem hiding this comment.
Indeed. I didn't realize that this uses /trigger instead of /build . The letter has a project wide rebuild option.
I wonder if we somehow can give the bot access to that route.
There was a problem hiding this comment.
openSUSE/open-build-service#2094 is the history of this route
There was a problem hiding this comment.
and openSUSE/open-build-service#15119 is a new attempt at getting a fix in.
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
6adbac2
to
a42d180
Compare
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
a42d180
to
22caeb0
Compare
dirkmueller
changed the title
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
22caeb0
to
681b988
Compare
| @@ -152,8 +174,16 @@ def run(self, version, token=None): | |||
| self.logger.info('No positive result from openQA (yet)') | |||
| return | |||
|
|
|||
| if rebuild_token: | |||
| for pkg in self.fetch_package_names_in_project(_DEVEL_PRJ_APIURL, devel_prj): | |||
There was a problem hiding this comment.
jftr OBS team will fix this that we can use unscoped tokens to trigger project wide rebuilds
There was a problem hiding this comment.
So shall we wait for them to implement this before moving this forward?
There was a problem hiding this comment.
implementation is ready at openSUSE/open-build-service#15119
There was a problem hiding this comment.
So shall we wait for them to implement this before moving this forward?
no, I think we can do this optimisation to use only one project wide trigger after merging this. however I'd really like to get one repo published first before adding more risk of regressions..
|
This will need a rebase as #3026 got merged. Technically the docstring for |
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
681b988
to
64ff801
Compare
Both done |
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
64ff801
to
b995451
Compare
and correct the docstring of is_repo_published
Unlikely, but who knows ¯\_(ツ)_/¯
The current version is iterating twice over the package list, but that is not necessary. The list can be assembled in a single pass, making it more obvious what is happening.
dcermak
force-pushed
the
bci_repo_trigger_build
branch
from
b995451
to
de1d827
Compare
|
openSUSE/open-build-service#15261 is now merged, I guess we could switch to it and test this (and get it over the hump)? |
| @@ -157,8 +182,16 @@ def run(self, version, token=None): | |||
| return | |||
| openqa_passed_packages.append(pkg) | |||
|
|
|||
| if rebuild_token: | |||
| for pkg in self.fetch_package_names_in_project(_DEVEL_PRJ_APIURL, devel_prj): | |||
| url = makeurl(_DEVEL_PRJ_APIURL, ['trigger', 'rebuild'], {'project': devel_prj, 'package': pkg}) | |||
There was a problem hiding this comment.
this could now be changed based on openSUSE/open-build-service#15261 (which is deployed)
currently untested