rescan headers on large newsgroups

[sikuraktur]

I'm subscribed to a number of large binary news groups.
If I have downloaded, deleted, and then purged most of the headers from one of these groups and then try to do a "rescan headers", the CPU goes to 100% but 0.000KB/sec bandwidth is consumed. (This only happens with "rescan", Never with "get new")
Newspro is still responsive to user actions, but selecting the header thread and pressing the "cancel task" button fails to stop it.
Selecting cancel all tasks also fails to stop it.
After leaving it in this state for several hours I give up and try to exit.
Trying to exit fails with the message: "Shutdown in progess. Processing remaining new headers or flushing database. Try emergency exit"
Trying to perform an emergency exit doesn't work either.
Eventually I have to resort to killing the process from task manager, then restarting and restoring the database.

This is reasonably consistent behaviour and I can reproduce it almost at will.
I'm running 3.8 but this was also happening in an earlier version.

Any ideas?

[alex]

i looked at the code and i'll publish a faster implementation here later (tonight?) in this thread, i'll replace the sorting algorithm, just check it works, i'm currently working on different aspects.

[alex]

here is the file:

http://www.netwu.com/newspro/npr38rscn.zip

has the problem gone?

actually a faster algorithm was at hand at other places in the code, just noone pointed out the problem before, if it is it.

[sikuraktur]

thx
i'll probably won't be able to test it 'til tomorrow

[sikuraktur]

alex
thank you for the quick response
apologies for not getting back to you sooner myself.
this change has cured the problem
thx again
sikuratur

[brianw]

I too have tried this and it fixed my problem as well.

Thanks alot.

[brianw]

I too have tried this and it fixed my problem as well.

Thanks alot.