On 29/07/2018 15:21, Jeremy Kloth wrote:
On Sun, Jul 29, 2018 at 3:13 AM Tim Golden
wrote: For an example:
Thank you! After inspecting all the errors, it does seem that they are ALL caused by "bare" os.unlink/rmdir calls. So it seems that a massive undertaking of ferreting out these locations and replacing them with their support equivalents would be needed to have a passing test suite for you.
Thanks for checking. As I mentioned elsewhere in this thread, I propose to go through and apply support.unlink where necessary. It won't make things any worse and will hopefully improve in some areas. For test_mailbox I've experimentally implemented a hybrid tempfile / local directory solution. ie I've created a new file on each run, but only within the python_<pid> folder which already exists. As long as the directory cleans up there should be no leftovers. That's certainly helped although my re-run harness has provoked at least one error. TJG