ehh...i'm still against it. if it were something we were expecting end-users to fix, i would agree. but i would prefer that we fail loud and large, because i don't think this it actually is an end user-facing issue.
if we add the logic to VersioncontrolBackend that we discussed yesterday, then the primary responsibility for specifying the default plugin rests on the backend implementation, not on the end-user. given that developers are the audience there, it's better that we have a hard failure (watchdog + exception thrown) than a soft failure (defaulting to a non-functional plugin), as failing to specify a default reposync plugin means that your backend really hasn't yet fully conformed to the interface.
Comments
Comment #1
sdboyer commentedehh...i'm still against it. if it were something we were expecting end-users to fix, i would agree. but i would prefer that we fail loud and large, because i don't think this it actually is an end user-facing issue.
if we add the logic to
VersioncontrolBackendthat we discussed yesterday, then the primary responsibility for specifying the default plugin rests on the backend implementation, not on the end-user. given that developers are the audience there, it's better that we have a hard failure (watchdog + exception thrown) than a soft failure (defaulting to a non-functional plugin), as failing to specify a default reposync plugin means that your backend really hasn't yet fully conformed to the interface.Comment #2
marvil07 commentedChanging target then.
Comment #3
sdboyer commenteddiscussed with marvil07, he agrees, so marking won't fix
Comment #4
marvil07 commentedxpost here I think, we still want one reposync plugin to use on tests.
Comment #5
sdboyer commentedthis is an important part of our test suite in terms of being architecturally complete, but we really don't NEED it to launch.