> > Intuitively, I agree. However, we have some assumptions in our test > > suite that if an authority is disabled, we remove its catalog entries > > unless a refresh is not allowed. In that case, we leave the catalog in > > place until such a time that we can connect to the network and refresh > > the catalog. I wrote a similar version to what's above, but found it > > didn't pass test. > > > > The other possibility is to change the test case, if we think it's > > incorrect. > > The test seems incorrect to me. If whoever wrote the test can justify it, > I'm open to being persuaded. Otherwise, please change the test.
I had that backwards. The delayed refresh case is failing, but it's because we change the origin for an authority and specify --no-refresh. In the old code, the catalog wouldn't get deleted but a full refresh for that authority would occur next time we executed an image-modifying operation. Now, the catalog gets deleted, but since we don't refresh the test fails, because it looks like the catalog is empty. The previous behavior was to let the stale catalog persist until refresh, but the current behavior is to remove the catalog as soon as it becomes out of date. Either seems correct to me, but we need to pick one. -j _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
