Re: [Pan-users] What changed view of headers?

[Duncan]

Maurice posted on Thu, 25 Sep 2014 17:16:30 +0000 as excerpted:

> A few months ago I suddenly lost all the header lists: Nothing appeared
> there. Empty for all subscribed newsgroups.
> 
> Checking View/header list I found that "Match only unread articles"
> was ticked. Unticked it and headers re-appeared.
> 
> Today the same thing has happened. Anyone had this? Known bug? What is
> ticking that option?!

FWIW I run with unread-articles only by default.  However, I have a 
custom hotkey (r) mapped to toggle the setting, so if someone only short-
quotes an article I want to see more of or if I need the larger thread 
context in general, I can hit "r" and have the full thread to browse. 
Then another "r" when I'm done switches back to unread-only mode. =:^)

But that's a custom hotkey, pan doesn't have one mapped to that function 
by default, so it can't be mistakenly hitting that.

But there's a toolbar button for it.  It's possible you're accidentally 
hitting the toolbar button.

The other possibility I know of for things like that, is filesystem 
irregularities.  If the filesystem pan (in this case, it can be other 
things, back when I ran MS Windows, naturally on FAT, having settings 
revert like that wasn't unusual; it's rather more unusual but not unheard 
of on Linux) is on isn't cleanly unmounted, perhaps due to a crash or to 
a suspend/hibernate that doesn't come back up cleanly, requiring a 
reboot, the log replay and/or fsck process can revert files to a previous 
condition or occasionally disappear them entirely.  If an affected file 
happens to be the one that setting is stored in...


Of course that's assuming you're not simply running into a pan (or 
library, gtk, etc) bug.

I know of one similar bug that sometimes affects me, in this case 
apparently a gtk bug.  Apparently there's some sort of race condition in 
the startup logic, such that if a particular pan/gtk resource (I believe 
it's the icons for the state and/or action columns in the headers pane) 
isn't loaded by the time pan needs to draw the header pane, that resource 
will come up null and pan will reduce to 1 the width (normally the 
standard 32 pixel icon size) of that column. Unfortunately, when that 
happens it screws up all the columns to the right as well, and I end up 
with the last column, score in my configuration, taking up the FULL 
HEADER PANE WIDTH!

The first time it happened it took me awhile to figure out, but 
eventually I did, and I painstakingly reset all the columns to the widths 
I wanted manually, fighting a buggy pan/gtk to get it to work.

After about the second or third time -- it seems to trigger after I've 
updated something, but that's as close to a specific reproducer as I've 
gotten -- I figured out which pan settings file (preferences.xml) was 
affected, and saved a copy of it that I could restore when necessary, 
thus avoiding all the fiddly column reconfiguring in the GUI.

Sometime later even that got tiresome, and I eventually came up with 
config-file patches that patch columns set to 1, back to their normal 
size.  I was already using a wrapper script to do some other setup before 
starting pan, so I added a patch routine to it, such that if the columns 
settings are normal it leaves them alone, but if I load pan and find them 
all reset to 1 again, I can simply restart pan, and the patches will 
patch the config file back to the normal settings.  So now, when I start 
pan and see the screwed up header pane, I simply restart pan again and 
the wrapper script reapplies the settings I want, so I'm back in business.


It's possible you found a similar bug, tho here I'd be unlikely to notice 
it since I do unread-only by default.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman