fix branch -N #1527
fix branch -N #1527
Conversation
Branching with missingok must not ask for existing package, it will lead to an error
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #1527 +/- ##
==========================================
- Coverage 31.48% 31.47% -0.01%
==========================================
Files 103 103
Lines 19072 19074 +2
==========================================
Hits 6004 6004
- Misses 13068 13070 +2 ☔ View full report in Codecov by Sentry. |
| @@ -6106,27 +6106,28 @@ def branch_pkg( | |||
| # BEGIN: Error out on branching scmsync packages; this should be properly handled in the API | |||
|
|
|||
| # read src_package meta | |||
There was a problem hiding this comment.
The comment should be moved with the code.
| if devel_project: | ||
| raise oscerr.PackageError(devel_project, devel_package, msg) | ||
| raise oscerr.PackageError(src_project, src_package, msg) | ||
| if not missingok: |
There was a problem hiding this comment.
Could you explain why we don't want to read devel project/package when missingok is on?
Don't we want to resolve the source project/package every time?
There was a problem hiding this comment.
-N is to prepare a new package for submission. It flags the _link file that the target is actually not there. So reading the meta from the target will always fail.
There was a problem hiding this comment.
I think this condition should be if not newinstance:
Also, shouldn't we explicitly turn off branching from the devel package when newinstance is set?
query = {'cmd': 'branch'}
- if nodevelproject:
+ if nodevelproject or newinstance:
+ # osc branch -N/--new-package should ignore devel project and always branch from the given package
query['ignoredevel'] = '1'
There was a problem hiding this comment.
well, it does not really matter as there could be no package definining a devel area. I would have kept the api call simple and not added that parameter in that situation.
There was a problem hiding this comment.
after some more discussion with Daniel I actually agree that ignoredevel makes sense when -Nis used in the situation when a new instance of a package is wanted in a hierachial project connected via project links (like SUSE:SLE-SP?: setups)
|
I assume this explains why |
|
@glaubitz yes, indeed. |
|
Closing in favor of #1535 |
Branching with missingok must not ask for existing package, it will lead to an error