PR#1701: Fixes for allow_dashes_in_svc enforcement - fm-orchestrator

fm-orchestrator

#1701 Fixes for allow_dashes_in_svc enforcement

Merged 2 years ago by mikem. Opened 2 years ago by mikem.

mikem/fm-orchestrator dashes-part-deux into master

defer commit of module entry until we set the context and validate

Mike McLean • 2 years ago

a8adfd4

use scm_stream instead of scm.branch

Mike McLean • 2 years ago

872320a

module_build_service/common/models.py

file modified

+3 -1

		`@@ -621,6 +621,7 @@`
		`rebuild_strategy=None,`
		`scratch=False,`
		`srpms=None,`
		`+ commit=True,`
		`**kwargs`
		`):`
		`now = datetime.utcnow()`
		`@@ -650,7 +651,8 @@`
		`module.buildrequires.append(base_module)`

		`db_session.add(module)`
		`- db_session.commit()`
		`+ if commit:`
		`+ db_session.commit()`
		`return module`

		`def transition(self, db_session, conf, state, state_reason=None, failure_type="unspec"):`

module_build_service/common/submit.py

file modified

+2 -2

		`@@ -102,11 +102,11 @@`
		`if isinstance(stream_or_packager, Modulemd.ModuleStream):`
		`# this is a ModuleStreamV2 object`
		`stream_or_packager = stream_or_packager.copy(stream_or_packager.get_module_name(),`
		`- scm.branch)`
		`+ scm_stream)`
		`else:`
		`# this is a PackagerV3 object`
		`stream_or_packager = stream_or_packager.copy()`
		`- stream_or_packager.set_stream_name(scm.branch)`
		`+ stream_or_packager.set_stream_name(scm_stream)`

		`# If the version is in the modulemd, throw an exception since the version`
		`# since the version is generated by MBS`

module_build_service/web/submit.py

file modified

		`@@ -701,6 +701,8 @@`
		`reused_module_id=params.get("reuse_components_from"),`
		`scratch=params.get("scratch"),`
		`srpms=params.get("srpms"),`
		`+ # don't commit now, we do it ourselves below`
		`+ commit=False,`
		`)`
		`module.build_context, module.runtime_context, module.context, \`
		`module.build_context_no_bms = module.contexts_from_mmd(module.modulemd)`

mikem commented 2 years ago

This is a follow up to #1700, which doesn't work as expected. There are two fixes here.

First, we actually use the scm_stream value that the previous patch sets, instead of scm.branch. This was an oversight in #1700

Second, we defer the commit of the module build entry until after we set the context and check the scv values. This was arguably a problem before now, but the checks added in #1700 exacerbated the situation by causing a failure before the correct context was set, resulting in a stray entry with the default context value of 0000000. This stray entry caused successive build submissions to fail.