pan-users
[Top][All Lists]
Advanced

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

[Pan-users] Re: Test patch: faster decoding


From: Duncan
Subject: [Pan-users] Re: Test patch: faster decoding
Date: Tue, 25 Apr 2006 10:12:16 -0700
User-agent: Pan/0.14.2.91 (As She Crawled Across the Table)

boo posted <address@hidden>,
excerpted below,  on Tue, 25 Apr 2006 15:05:33 +0200:

> i also noticed that some 50mb files seem to not get decoded at all (no
> gui-freeze either), just like the issue with the incomplete files, but
> those files arent marked incomplete. all the parts get downloaded to
> ~/.pan2/article-cache, but just stay there, also there's no entry in the
> events-window. i'm not sure that this is because of the size, but it
> only happend on those for me, on like half the files i tried to download
> though. i then tried to download the tasks.nzb with klibido and it went
> fine, meaning no corrupt parts or anything like that, that could
> interfere with the decoding.

This may be related to the same bug I'm seeing on text posts, where they
show up in the overviews/headers pane, I can click on them and see the
download, but PAN doesn't mark them as downloaded, and doesn't display
them in the body pane either.

What's strange is that even some of my own posts, made with pan 0.14.x,
exhibit this behavior.

I've been sick the last couple days, spending a lot of time horizontal,
and when I'm vertical, my fevered and throbbing head hasn't wanted to try
to tackle the issue.  I do know, however, that it's in both 0.93 and 0.94,
and believe it was in 0.92 as well, tho I barely tested it as it as there
were enough other issues with it that the test didn't go far.

With the quote-color options back in 0.94, so I can actually read the
posts that do display, this is the big bug remaining before I think it'll
be usable enough to do most of my news in, thus testing much more
thoroughly.

Charles, it looks like all the  --debug options that were in 0.14 and
previous are gone in 0.9x.  Is the only way to debug using a tracer, now?

I routinely compile with gcc-4.1.0, using these CXXFLAGS (CFLAGS are
similar, but with -freorder-blocks-and-partition as well, which doesn't
work so well on C++, thus the difference, -ftree-pre is new to gcc4, so
doesn't work on gcc3):

CXXFLAGS="-march=k8 -Os -pipe -fomit-frame-pointer -frename-registers
-funit-at-a-time -fweb -freorder-blocks -ftree-pre -fmerge-all-constants"

CFLAGS are similar, but with -freorder-blocks-and-partition as well, which
doesn't work so well on C++, thus the difference. -ftree-pre is new to
gcc4, so doesn't work on gcc3.  Yes, I know some of those
(-fomit-frame-pointer on amd64, it wouldn't be on x86, and
-funit-at-a-time) are redundant, but they are there in case a (Gentoo)
ebuild replace-flags the -Os to something else, as some do.

When I get feeling good enough to pursue it, I've been planning to
recompile with simpler flags, and/or with gcc 3.4.6, to see if that
changes the results.  Grabbing the cached post (if possible), to compare
with what 0.14's version looks like, and posting it here (the posts have
been in the ISP-private cox.* heirarchy, so no posting a public
reference), are also on the list, as is finding examples on a public
group, for live testing by others.

-- 
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 in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html






reply via email to

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