Code review comment for lp:~vila/bzr/1451448-skip-racy-tests

Thanks for the review.

> Thanks for the fixes, Vincent. Both situations seem suspect:

> running the http client and server in the same thread seems prone to races,

Historically, all client/server tests has been written this way since day one. It took years to run into the races and realize the root cause was that the python library was not intended to support both the client and the server in this context.

So, after the fact, it seems obvious that this was a bad design decision but the thing is: when everything works and once in a blue moon a test fails, nobody suspects that design... ;)

> running
> windows filesystem-specific tests on a non-windows filesystem sounds like it
> would be likely to fail.

Well, history again, it was easier at one point to be able to run those tests on ubuntu to make sure the windows support wasn't regressing until tests can be run there. Nowadays... it's more complicated.

Hope this shed some light on the past ;)