Re: [Pan-users] One group suddenly ignored

[Duncan]

Glen Walpert posted on Wed, 13 Apr 2011 18:21:50 -0400 as excerpted:

> <clip previous discussion of problem with older versions>
> 
> I have now installed pan 0.134 on two machines, one maintaining my old
> configuration (.pan2 directory) and one clean install where I manually
> reconfigured my server (news.astraweb.com) and added several newsgroups
> including the problem group (sci.electronics.com) but no related (likely
> to have crossposts) groups.  Get new headers on startup and when
> entering groups is disabled in preferences.  Delete old messages set to
> 3 months.  Symptoms are the same on both systems:

(Separating these with vertical space so pan's wrapping code won't 
butcher the wrapping so badly...)

>   - groups other than sci.electronics.design work normally

>   - I can get and look at new headers in sci.electronics.design,
>     and read the messages, with:

>        - get new headers, which actually gets only those headers
>          actually new since I last checked

>        - get headers for last N days - get all headers - get last N
>          headers

> All of these appear to work normally, however, as soon as I leave
> sci.electronics.design and return, they all vanish!  If the new headers
> are retrieved with "get new headers in subscribed groups" or by having
> get new headers on startup enabled, then the number of new messages is
> shown in the groups pane but when the group is entered they all
> disappear, and regardless of how headers are retrieved the headers file
> (.pan2/groups/sci.electronics.design) is not updated!

That was back in April.  I had the post saved as unread, to follow up on 
later, when I had something useful to say about it, as the problem is 
similar to one I had (yes, *HAD*, see below) for quite some time.

A lot has happened since then.  0.135 is out, and Heinrich Mueller 
appeared and shook up the pan world with a number of very interesting new 
features, some of which we've been waiting on for a VERY long time (a 
decade more more, for good binary posting).  While these features aren't 
stable yet and may take a bit to make it into a release version, the fact 
that we have them at all, particularly binary posting which as I said, 
pan has been waiting on for at least a decade, is HUGE news! =:^)

Of course, HMueller has made some other less "headline" changes in his 
github sourced version as well.

One of those... I /think/ but am not /sure/, was that somewhere along the 
line, several formerly "pan-bug-dead" groups that I had subscribed to on 
gmane SUDDENLY came back to life again!

Since AFAIK the bug was never fully traced, I can't say for sure whether 
it was one of those small changes by HMueller that did it, when he 
noticed a logic bug and simply fixed it, OR whether it was changes on the 
gmane side, as it had some problems for a few days, but came back from 
them, and it was shortly after that, that those groups simply "unbroke", 
here.

What's weird, tho, is that they didn't all "unbreak" at once, at least 
that I could see.  They came back one by one over about a week's period.  
I /think/ that might be explained as the first post to the now "unbroken" 
group triggering pan to rethread it or whatever, and since some of those 
groups/lists are rather low traffic, they didn't show up as unbroken 
immediately, only after a post.  I DO know that one of the higher traffic 
ones was the first one I noticed that it had "come alive" again.

Anyway, whatever the reason, the formerly broken groups DO seem to be 
fixed, here.  Please post whether sci.electronics.com has "unbroken" for 
you or not, and which version of pan you are using.  If you're using a 
mainstream version and it's still broken, could you possibly try 
hmueller's git-repo version, master branch, and see if it works for you?  
Note that depending on how active that group is, it may take several days 
for it to visibly "unbreak", as it did for some groups (lists but seen as 
groups on gmane) here.

If you can confirm a fix, then perhaps we can git-bisect it (or with 
hmueller's help simply git-cherrypick and test) to a specific commit, and 
we can have khaley apply that to his master for further testing.  If that 
too works, pkovar can integrate it into the main upstream gnome repo, and 
0.136 or whatever, can contain the fix.  That bug has been vexing enough, 
at least personally, that I'd argue if we can verify a specific fix, that 
alone is worth a quick official version bump.

Of course, at this point it's just as possible that the problem was fixed 
by the gmane server-side reindexing that the probably had to do after 
whatever issue they had.  (From the post to gmane.discuss, it was due to 
a misconfigured log using up all the space on the disk, but while they 
were fixing that, I think they made a couple other minor adjustments, 
possibly fixing the indexing or whatever issue it was that had pan bugged 
out.)  But even confirming that for sure would be useful.

Unfortunately, now that it's finally working, here, for which I'm VERY 
thrilled (!!! =:^), I'm rather less than thrilled at the prospect of 
testing to try to figure out what fixed it, potentially breaking things 
once again.  Now that it's working, I'd *MUCH* rather leave well enough 
alone.  Thus, first seeing if it "magically" starts working again for 
you, either that it already did, or that it does when you switch to 
current hmueller (imhotep82 on github) master, seems reasonable.  Then 
perhaps, with good backups and in the interest of science...

I've actually been rather reluctant to post about it, too, for fear I 
might jinx things! =:^(  But it has been working fine now for a week or 
10 days, so here's hoping I haven't! =:^)

-- 
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