pan-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Pan-users] problem with 0.13.94


From: Jim Reiss
Subject: Re: [Pan-users] problem with 0.13.94
Date: Mon, 14 Apr 2003 15:15:47 -0600 (MDT)

> >>    FYI love the new release of pan, but I'm having a problem with it.
> >>    More often now then ever before, when switching around large binary 
> >> groups, I've noticed that pan will just abnormally close. It doesn't 
> >> even report a seg fault or anything, just poop gone. I was thinking it 
> >> might be that while I'm loading a group, I click to try and load 
> >> something else, and that crimps it. Anyway, just wanted to get the idea 
> >> out there that its doing it more so now then in the last minor release.
> 
>   Pan crashed!  I'm not using it      There's a crasher in one of Pan's 78,000
>   again until it's fixed!             lines.  I want you to spend the next
>                                       week guessing which one.
My observations, in case they might help narrow down the crashing and
leaking issues a little from that 78,000:

1. Pan just uses a lot of memory to load all the headers for a large
group, and by default it will download headers from several groups
simultaneously (any maximum number of tasks above 1).  On many systems
the amount of memory used is so great that pan will run out.  If there
was a configuration option to download headers from only one group at a
time, this problem likely would go away for most users.

2. The article database appears to get corrupted if you're downloading
headers for the same group from more than one server.  Even if the
solution suggested in #1 was implemented there would still need to be a
mechanism to make sure the same group is not being downloaded in more
than one task (regardless of server).

3. The new crashes I've seen in recent versions seem to be related to
multithreading.  Specifically, I click on a group and it starts loading,
but pan also starts reapplying the filters or "updating header pane" or
somesuch in the group I'm leaving.  The new group finishes loading and
then the other thread finishes doing reindexing of the no-longer-current
group, I see a flash of the headers from the old group, and then crash.
It seems to me that it's related to how pan assumes something has been
entered in an input field even when you haven't pressed the enter key
(e.g. I type the name of a group and then switch to "All Groups" from
"Subscribed Groups", and it applies the name as a filter even though I
didn't press enter).  A possible fix would be to create a mutex for each
server that would allow only one thread to work on header loading/sorting/
filtering at a time.

I don't know if these are enough information to help, but maybe in
conjunction with other feedback they will spark an idea.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]