GNOME Bugzilla – Bug 86795
"423 No such article" should not be retried
Last modified: 2004-12-22 21:47:04 UTC
I suggest that if the reason an article could not be retrieved is a message like, "423 No such article number", Pan should immediately fail that task, instead of retrying it. Unlike a socket error, authentication, or too many connections, a 423 is not likely to be resolved by retrying. Furthermore, Pan's behavior if several articles are trying to be read but return 423s is quite annoying -- each time an article returns an error message, the article pane jumps to that article and we get to see the error printed in the article space. A red X has already been marked by its header the *first* time the 423 came back, so I know it has failed. I'm not familiar with NNTP, but is it possible 4xx errors mean, "It's impossible to give you this article at all" whereas 5xx errors mean, "I can't give it to you for some reason"? (I seem to recall SMTP errors are organized in such a way.) If so, then perhaps we should give up when we get any 4xx.
AFAIK, Pan doesn't retry when it gets a 423 error message. I've tried this, both by reading and by saving, and don't see a retry. Why do you think it retries ? Can you give me a recipe to reproduce this?
Created attachment 9696 [details] Debug runlog of retry behavior
I think it is retrying because I can observe it doing so. The runlog I just attached shows it trying six times to grab a message that was marked with a red "X" the first time it tried, because server reported 423. The next message I selected (follow-up to the first one) was present, so it was displayed normally. Here, the "recipe" to reproduce just seems to be, "Try to read an article that is no longer present on the server."
I've now managed to reproduce this, but only by switching off 'Automatically remove failed tasks from the task manager'. (this is why I just love behavioural preferences :)) When that is off, and I read a message that's not on the server, I see retries in the task manager and the console, and I get a number of '423' error messages in the text window. The number of messages is related to the number of retries (e.g. four messages when retry is set to three). Can you confirm you're experiencing the same behaviour and circumstances? I'll look into fixing this tomorrow night.
Fixed for 0.12.1 and merged to HEAD.