Of course the real point is these slipped through because of the lack of tests that exercise them…

Actually I though about if I could write tests for that, but I don't have a good idea how one could do that, because this is e.g. the exception handling for the writing of data within the spew method, so unless you want to refactor the code in some fancy way I'm afraid you'd have to come up with a test where the open-ing of the file works but then the print fails, and I can only imagine that happening with a full disk.

Thanks!

I'm afraid you'd have to come up with a test where the open-ing of the file works but then the print fails, and I can only imagine that happening with a full disk.

Well, it depends. Some of the I/O operations are already written using some method call on a file handle. If the prints and closes etc. were also written that way then all that would be necessary to exercise those branches would be for the filehandle method to return an object with corresponding methods that return the appropriate errors. Or I think you could do it without touching the code by having filehandle return a tied handle that isn’t backed by an actual file handle, though that is slightly fiddlier.

@xdg Any interest in any of that?

No, I think that sort of error path testing is more effort than it's worth.