Marc.theaimsgroup.com has a lovely "raw mode", but it truncates posts that are
too long.
groups.google.com looses attachments, even plain text ones.
Anything based on hypermail screws up whitespace in an interesting way. (You
can't "save as" because it inserts <br> between each line, even the source
preserves whitespace and keeps tabs as tabs and everything. You can't cut
and paste because html that isn't quoted with <pre> mangles whitespace for
you. You can't make a simple script to remove <br> when it occurs at the end
of the line because it's not consistent: blank lines don't become <br>, they
become <p>. And you'd have to worry about < and such anyway...
I'm sorry, I'm a bit frustrated right now. I"m trying to provide URLs to
patches posted on this list. This sounds easy, doesn't it? It's not.
I'm going to go to lunch now...
Erich: Could you put your october 25 numa scheduler posting on your home page
somewhere? I tried:
http://home.arcor.de/efocht/patches/01-numa_sched_core-2.5.44-10a.patch
But that just would have been too easy... :)
--
http://penguicon.sf.net - Terry Pratchett, Eric Raymond, Pete Abrams, Illiad,
CmdrTaco, liquid nitrogen ice cream, and caffienated jello. Well why not?
On Sunday 27 October 2002 17:01, Rob Landley wrote:
> Erich: Could you put your october 25 numa scheduler posting on your home
> page somewhere? I tried:
>
> http://home.arcor.de/efocht/patches/01-numa_sched_core-2.5.44-10a.patch
>
> But that just would have been too easy... :)
Sorry that I didn't do this before...
Patches are at
http://home.arcor.de/efocht/sched/01-numa_sched_core-2.5.44-10a.patch
http://home.arcor.de/efocht/sched/02-numa_sched_ilb-2.5.44-10.patch
Regards,
Erich
On 2002-10-27, Rob Landley <[email protected]> wrote:
> Marc.theaimsgroup.com has a lovely "raw mode", but it truncates posts
> that are too long.
This is true, and I consider it a bug. However the limit should be quite
high--either 256KB or 1MB or so. So raising / eliminating the limit isn't
high on my priority list :-P Please do let us know (me directly, or
[email protected]) if you find other mails that have been truncated
at shorter than this, I'll look into it.
Also, I'm interested in any corner cases where the attachment-parser messes
up--most of all when it fails to make attachments properly downloadable,
but also, to a lesser extent, any predictably readable mime-type, encoding,
etc which it currently doesn't try to print in-line, but could.
> Erich: Could you put your october 25 numa scheduler posting on your
> home page somewhere? I tried:
> http://home.arcor.de/efocht/patches/01-numa_sched_core-2.5.44-10a.patch
If I've got the right mail, you're after this one, correct?
http://marc.theaimsgroup.com/?l=linux-kernel&m=103556765829589&w=2
...Which looks complete from here? That particular patch can be downloaded
directly as:
http://marc.theaimsgroup.com/?l=linux-kernel&m=103556765829589&q=p3
Thanks,
Hank Leininger <[email protected]>
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Sun, 27 Oct 2002, Rob Landley wrote:
> On Sunday 27 October 2002 22:34, Hank Leininger wrote:
> > Also, I'm interested in any corner cases where the attachment-parser messes
> > up--most of all when it fails to make attachments properly downloadable,
> > but also, to a lesser extent, any predictably readable mime-type, encoding,
> > etc which it currently doesn't try to print in-line, but could.
>
> This one went totally bonkers, but the post it was embedded in could have been
> too long:
>
> http://marc.theaimsgroup.com/?l=linux-kernel&m=103559201620433&q=p3
Hm. Actually I think the only problem with this is a browser issue.
[ But not any more, see below. ] That attachment is named
"linux-2.4.20-pre9-agp3.patch.gz". Netscape (at least) has absolutely
idiotic behavior when downloading .gz files: it will strip the .gz
extension, but not decompress the file! So when I use Netscape to save
the above, I get:
- -rw-r--r-- 1 hlein users 7023 Oct 28 02:06 linux-2.4.20-pre9-agp3.patch
Which is a bunch of binary crap. But 'file' realizes it's compressed:
$ file linux-2.4.20-pre9-agp3.patch
linux-2.4.20-pre9-agp3.patch: gzip compressed data, deflated, original filename, `linux-2.4.20-pre9-agp3.patch', last modified: Thu Oct 24 19:09:39 2002, os: Unix
And if you rename it to put the .gz back on, and/or gunzip -c it, you
should get what looks like a proper diff file. (I don't know a way for
a server to override this behavior; if anyone does, please contact me
off-list.)
Hm. OTOH, I just tested that with Konqueror (v 3.mumble) and it
segfaults trying to download it. Bad.
[tests]
OK. I was basically reproducing the Content-Disposition header from the
mail, which contains a filename hint. In an email it looks like:
Content-Disposition: attachment; filename="linux-2.4.20-pre9-agp3.patch.gz"
Netscape and at least some other browsers eat that just fine (and take
the filename hint). Konqueror segfaults. If I remove that header, it's
fine, but then there's no filename hint. If I s/attachment;// then
Konqueror doesn't segfault. Netscape 4.8 and MSIE 6sp1 both still grok
the filename hint. Konqueror's handling of the filename hint seems
inconsistent (sometimes doesn't work at all; sometimes thinks the
closing double quote is part of the filename). Still investigating.
But this is better than before, so just committed that change.
I'll file a bug report on Konqueror (the header was admittedly bogus
before, but it shouldn't have caused a SEGV), and try to figure out how
to get filename hints working better for Konq as well.
> linux-kernel mailing list blurb was at the end of them. The penultimate list
> will be posted in a few minutes, just trying to catch up on linux-kernel
> first to see if I missed anything.
If these mails don't cross each other, please check if you can if the
other cases you've seen are similar to the above.
Thanks,
Hank Leininger <[email protected]>
E407 AEF4 761E D39C D401 D4F4 22F8 EF11 861A A6F1
-----BEGIN PGP SIGNATURE-----
iD8DBQE9vOwMIvjvEYYapvERAhZwAJ4pNssBzCb5rRnt3A9t3hDp0od6FwCffStm
cUM7WhDPAGdaQSgyNecGMU0=
=6Z+U
-----END PGP SIGNATURE-----
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Sun, 27 Oct 2002, Hank Leininger wrote:
> high on my priority list :-P Please do let us know (me directly, or
> [email protected]) if you find other mails that have been truncated
> at shorter than this, I'll look into it.
Thanks to Nicholas Wourms and a few others, it has been confirmed that
I am, in fact, an idiot (for some, there was never any doubt...).
Somewhere along the way the message bodies stored in MARC got assigned a
64kb max size. So messages were indeed truncated at 65,535 bytes.
That's fixed. Out of 3.1 million mails MARC got in 2002, 7,184 had been
silently truncated. I've identified and fixed from backups the majority
of those (and 2001 as well); there are still some 1,387 mails from 2002
which appear truncated. So we're down from 0.23% bad data to 0.04%. I
believe the majority of those are from bulk inserts I've done of new
lists added this year, so lists like linux-kernel which we've carried
forever should be at or near 100%. And of course, all new incoming mail
should be fine, up to about 1MB.
So. *Now* please let me know if you find any mails to linux-kernel or
others that appear to have been truncated or otherwise damaged. ;)
Thanks,
Hank Leininger <[email protected]>
E407 AEF4 761E D39C D401 D4F4 22F8 EF11 861A A6F1
"This house... is clean."
-----BEGIN PGP SIGNATURE-----
iD8DBQE9wFUOIvjvEYYapvERAuCsAJ4+qG8cnG/WCkeB2kHzNAw68X1u3gCeNHQf
26+2zoHkJ1i1sy1wWunZVW0=
=SMQe
-----END PGP SIGNATURE-----