Redmine

Yes I think they should have because r1893 was committed on a "can't reproduce" basis and broke any possibilities to include a UTF-8 encoded parameter in the URL (was utf-8 RoR param in this case).
I don't see the point of pre-escaping a parameter which will be necessarily URL-escaped at the time the request is issued.
And I think that #1826 could have been fixed by using relative URLs which wouldn't have been processed by Apache mod_rewrite.

Could you please discuss them before reverting other people's commits?
It makes 2 times for reasons that could have been discussed before, it's rude, very frustrating and not really motivating.

Etienne Massip wrote:

Yes I think they should have

Sure but changing a test just to make it pass is not the way to go either. Updating and getting an error when logging in is pretty frustrating too :-(

Could you please discuss them before reverting other people's commits?

OK, sorry for that. Next time, let's discuss before actually starting to do the changes that should be discussed. Thanks for digging into this anyway Etienne.

Jean-Philippe Lang wrote:

Sure but changing a test just to make it pass is not the way to go either. Updating and getting an error when logging in is pretty frustrating too :-(

I did not change the test just to make it pass but because it was part of r1893 which I reverted; the after-login redirect was then expected to work just as it did before r1893 (or a test failure) but as you discovered it was not the case and I'm really sorry I broke it :(

Next time, let's discuss before actually starting to do the changes that should be discussed.

I did not expect any side-effect but as I updated common code I still added you as a watcher and didn't merge anything to branches or closed the issue.

As a regression, shouldn't it be fixed in 1.4 and 2.0 branches as well?