Conversation:
Notices
-
@imojito You don't change stuff in upstream libraries unless you want to get a violent migrane.
Also you might've missed: https://social.umeahackerspace.se/notice/334025
-
@another Hm, what were you trying to do?
-
@another Yeah, but what is causing it? Which script is executed when that occurs? As far as I know UnQueueManager only runs when trying to import feeds or when queues are _disabled_. The first of those is undoubtedly buggy, the second is not recommended.
-
And by import feeds I mean manually restoring accounts or bookmarks from externally stored files.
-
@imojito @another I think it's related to a switch from http to https at gnusocial.net and I'm going to make sure !gnusocial handles it better.
-
@another Chill out. It won't be fixed until the code's done. Nothing is magic. I'm working on it, just hang in there or I'll be put off and stop trying to solve the problem and you'll have to do it yourself.
-
@heluecht Yes, StatusNet sucks at it and that will never be fixed. Right now I'm writing the code to retry HTTPS for remote PuSH callbacks.
-
@imojito @anotheruser I don't think we should trust Location redirects anyway in this case.
-
@imojito Because if I do an HTTP request to you and get a Location: header back, that Location: header can be set by _anyone_ along the path... Never. Trust. HTTP.
-
@imojito HTTPS is more trustworthy of course, but our problem in 99% of the cases is that people switch from HTTP to HTTPS. If they change domain names or PuSH callback endpoints, that's an entirely different matter.
-
@imojito This is not a privacy issue (then we'd just get rid of HTTP altogether). It's a trust issue on when to update the URL callback. That POST->GET thing I have no idea how you concluded that it performs and I don't think it does that. It just drops dead on POST failure (note that the GET requests get 200 OK, but POST get 404 - and since our error message in the l…
-
@imojito Those logs have neither timestamps nor any kind of client identification. They tell me nothing.
Also, they're not to the same URI. So your point is entirely unvalid.
-
That looks better! Now, do we patch it?
-
I like spaghetti.
-
@imojito I think I did it already. https://git.gnu.io/gnu/gnu-social/commit/f4feef477b1a5a5cf5f3f79e6774bf047ad053b4
-
@imojito Otherwise, $request->setConfig(['strict_redirects'=>]); will do what you want, without having to rape and pillage the poor upstream library.
-
=>true of course.