Re: [Pan-users] Fetching new article headers in subscribed groups ever X minutes?

[Duncan]

Christian posted on Sat, 23 Feb 2013 17:21:55 +0000 as excerpted:

> Hi,
> 
> I am new to Pan. I'm running Pan 0.139. I have tried to figure out how
> to have Pan check for and fetch new article headers every 30 minutes,
> but I can't figure out how. In other news readers I've used it's been
> pretty obvious how to set this, but in Pan it's either hidden or I am
> missing something obvious. Please let me know how I can set it so that
> Pan fetches new articles by time interval, in my case 30 minutes.

This is interesting/distressing...

For some time, pan has (well, had, see below...) the feature of being 
able to download headers from the specified groups when given the 
appropriate command line interface (CLI) option, and as with any self-
respecting Unix-based program, one was expected to simply setup a cronjob 
from there, if scheduled activity was desired.  The CLI option was (based 
on the sources, which I just looked at) supposed to work like this:

pan headers:group.name1,another.group.name2,third.group.name3[,...]

... with the group names separated by commas if there was more than one.
There was (and remains) the --no-gui option, as well, to do it without 
popping up the GUI.

Unfortunately, somewhere along the line that feature apparently bitrotted 
and no longer works (just tested, I get a segfault), at least not with 
the 0.140-live-git-version I'm running, and based on your post, I'd 
assume not with 0.139 as well.  I don't remember anyone ever bringing up 
the problem here (tho it might have been bugged, I didn't check) and I 
hadn't done binaries at all for years until literally the last week or 
two, and didn't actually use that CLI feature (like most I guess I do 
mostly GUI) back when I did binaries, so I have absolutely NO idea when 
the feature bitrotted, but I do remember people using either that same 
feature or a very similar CLI-based-header-fetch feature many years ago, 
so I'm sure it must have worked back then.

But one of the coders (I'd guess either Heinrich or KHaley, depending on 
when the problem was caught) apparently became aware of the problem, and 
the --help output for the headers: option (and a couple others) is now 
commented out, with the comment "Doesn't work yet."

I'm almost positive the feature worked back half a decade ago or so when 
Charles was still actively developing pan, so that "yet" in the comment, 
instead of "currently broken" or similar, hints to me that the comment 
was added sometime post-Charles-era, and that whoever added it wasn't 
aware that it had worked in the past (assuming it did, but as I said, I'm 
almost positive it did...).

Which is all rather distressing to me, not only because of the broken 
header-fetch functionality itself, but because of its impact on the 
recently added "actions" feature, as well.  One of the big issues with 
pan, at least since the C++ rewrite, has been that it had a way to fetch 
headers (aka overviews) from the command-line, and thus could be setup in 
a cronjob or whatever for scheduled automatic header-fetches, but that it 
had no way of auto-fetching bodies (actual posts) as well.  The "actions" 
feature (see the actions tab of pan prefs) FINALLY allowed score-based-
auto-actions, including auto-cache and auto-save, and I thought with the 
appropriate actions and/or score settings and an appropriate cronjob, pan 
could FINALLY download both headers and bodies entirely unattended.

Unfortunately, if the CLI-based header-downloading feature is broken, so 
is the new actions feature, or at least a good part of the benefit it was 
supposed to bring! And here I was thinking that since I was back into 
binaries now, I might actually try out the feature!  =:^(


Hopefully Heinrich will jump in here with some comments about what might 
have gone wrong in the code and how easy it might be to fix, especially 
since it's likely him that commented the corresponding --help output (if 
I knew git a bit better I could use git blame to trace down exactly when 
that was commented, and maybe look further back to see when it broke, tho 
it might have been before pan switched to git from SVN, in which case it 
might be a dead-end search...), and it'll likely be him committing the 
fix.  But if it was trivial to fix, I'm sure it'd have been fixed instead 
of the associated --help output commented, so I'd guess it's at least a 
bit beyond "trivial". =:^(  OTOH, since the comment /does/ say "yet", 
it's likely that whoever commented that output never new it worked in the 
first place, so didn't know it was simply broken and thought it had never 
been implemented at all, so it's quite likely the overestimated the 
complexity of actually getting it working (again).


This was my Friday so I do have the next couple days off, but I really do 
need to do some house cleaning and other chores...  If I have time, 
however, or decide to be lazy and play on the computer rather than doing 
the cleaning I really SHOULD be doing, since I do build pan from git here 
and thus do have pan's git history back to when it was transferred from 
SVN, anyway, I might look into it further.  If I'm lucky, I'll find the 
commit that broke it as well as the commit that added the comment, which 
should definitely help in fixing the problem, and if I'm REALLY lucky, 
it'll be obvious what went wrong in that commit, even to this non-coder, 
and if I'm REALLY lucky, I'll even be able to come up with a fix.

We'll see...  But I REALLY do need to spend at least one day cleaning and 
doing laundry, etc, so who knows if I'll get to it at all, let alone get 
lucky enough to find the problem and have it be something simple enough I 
can actually fix. =:^(

Of course, should anyone with the skills have the time to look into it 
first and post a patch that I can simply apply and test... =:^)


Meanwhile... alternative that should be a functional workaround...

Consider installing and running a local news server such as leafnode, 
running on the localhost loopback interface (127.0.0.1 in IPv4, someday I 
need to jump in and learn IPv6...).  You can setup the server to do the 
scheduled pulls, then set pan up to connect to the local server over the 
localhost loopback.

-- 
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