[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Pan-users] Pan/0.139 - threading issue [solved - maybe]
From: |
Dave |
Subject: |
Re: [Pan-users] Pan/0.139 - threading issue [solved - maybe] |
Date: |
Fri, 27 Sep 2013 01:02:34 +0100 |
On Sunday 15 September 2013 14:03:34 Duncan wrote:
> Dave posted on Sat, 24 Aug 2013 21:37:30 +0100 as excerpted:
> > On Saturday 24 August 2013 11:28:15 Dave wrote:
> >> On Friday 23 August 2013 21:42:39 Heinrich Müller wrote:
> >> > I also strongly suspect it's the fault of either gmime or another
> >> > library. I won't investigate this in the near future, but you could
> >> > create a bug ticket for this issue.
> >> > I know I haven't been actively developing Pan in the last few months,
> >> > but I'll come back to it when I have the time.
> >>
> >> Hi Heinrich,
> >>
> >> Yes, I'll do that once I work out how :-)
> >> (and thanks for keeping a great newsreader up to date)
> >>
> >> I've also discovered that the "effect" changes slightly if I change the
> >> MID. The default "random" MID allows a deeper level of threading before
> >> the MID is split in the "wrong" place, depening on the length of other
> >> posters MIDs in the References: header. Using the custom domain MID
> >> which displays as pan. {date}:{domain}.{tld} causes the problem from
> >> the first reply.
> >
> > After trying and failing on FreeBSD and Ubuntu to build from latest Git
> > version, I deleted gmime26 (only Pan requires it), installed gmime24
> > and edit the FreeBSD port Makefile for Pan to look for gmime24 instead
> > of gmime26.
> >
> > Replies are now threading correctly and MIDs in the References: headers
> > are split correctly, ie NOT in the middle of the MID.
> >
> > Whether this is a gmime26 problem, a Pan problem or a FreeBSD ports
> > implementation of either gmime26 or Pan is beyond my level of
> > understanding, but at least it works now.
>
> Dave, do you remember or can you lookup what specific version of gmime
> 2.6 it was? Was it 2.6.17 perhaps?
>
> I had someone complain about my threading too, and I'm on gmime 2.6.17
> presently (on gentoo), but looking at what I have stored in binpkgs, I
> have 2.6.10, .13, .15, .16 and the currently installed .17; no gmime 2.4
> versions stored at all. And the build date on the oldest one, 2.6.10, is
> over a year ago, Aug 14 2012. I've no idea how long I was on 2.6 before
> that, but I'd guess it has been awhile.
>
> I can't be sure, but I'd have guessed the problem would have been
> reported /long/ before now if it was the full gmime 2.6 series, so I'm
> guessing it was 2.6.15 (March installation here) or later, very possibly
> only 2.6.17, which I built and installed on August 22 according to the
> date on my binpkg. You started this thread on August 19, it appears, so
> it's just possible you had just updated to 2.6.17 yourself at that
> point. I hadn't seen the problem yet then, but I had a complaint a few
> days later, and if I updated on the 22nd and it only happens on long
> threads where the references header is long enough to wrap, that could
> very well be it.
>
> So if you can double-check what version you were running into the problem
> with, it'll help narrow things down.
>
> I can actually either do some testing or look at the changelogs and/or
> commit logs and diffs, but narrowing down the problem as far as possible
> before I do will help. I guess I can do bug searches on it as well, both
> at gentoo and upstream.
Since I just rebuilt my system, Pan is now 0.139, it was previously 0.135 when
the problem first appeared. Current version of gmime is 2.6.18. Problem
appeared when I was using gmime 2.6.16.
I "fixed" it previously by reverting to gmime 2.4.x. (FreeBSD ports contains
both 2.4.x and 2.6.x,defaulting to 2.6.x if installing gmime without
specifying the version.
Your testing in the other thread indicates a change between gmime 2.6.15 and
2.6.16 which seem to narrow it down and confirm that is where the problem is.
(or maybe the way Pan uses/access gmime APIs??)
I'm sure other newsreaders must use gmime so I'd have thought if it was a
purely a gmime bug then other people would have encountered it before now.
Having said that, a quick grep through https://www.freebsd.org/ports/news.html
indicates Pan is the only one there using gmime.