oopses for failing code imports

[Jelmer Vernooij]

For svn and git repositories, the importds can now distinguish between remote repositories going missing, known bugs (nested tree support, filenames with backslashes) and other problems.

At the moment there are 34 failing svn imports and 9 failing git imports. I have some pending work that will bring the number of failing svn/git imports down to approximately a dozen in total.

It would be nice if new failures for svn/git imports triggered oopses. Code imports automatically get suspended after 5 failed imports, so this would only file OOPSes for new code imports.

[Robert Collins (lifeless)]

An OOPS is something a developer or sysadmin needs to look at, always.

I don't think thats the case with an arbitrary failing import: it
could be many different things such as a bad url, through to a code
bug.

If you can reliably separate out 'this is a code bug', sure, but
generating oopses for other reasons would be a step back in our
signal-to-noise ratio around OOPSes.

[Jelmer Vernooij (jelmer)]

I recently added code to the importds that filter out exceptions that indicate user errors - unable to contact host, branches that don't exist, etc. This works reliably for the existing imports.