On Tuesday 20 March 2007, Stefan G. Weichinger wrote:
>Gene Heskett schrieb:
>> Some history please, because you are not the only sufferer of this
>> problem. Send tar version, kernel version please.
>
>My report here:
>
>app-arch/tar-1.16-r2
>(stable)
Humm, may be suspect. I wish gentoo would not change the version numbers
rather willy-nilly so that those of us used to the gnu.org version
numbers applied to the tarballs (this IS the src of tar BTW) among the
various distro's. I've heard maybe 2 other similar queries, naming that
1.16 puppy.
># cat /proc/version
>Linux version 2.6.19-gentoo-r5
>(stable)
Once again, I wonder what the real patch level is that gentoo kernel is.
Its an arbitrary name that doesn't appear to follow the kernel.org naming
convention, so who knows how many patches may have been backported to it.
So go get, from kernel.org, the linux-2.6.20-tar.gz, and the
patch-2.6.20.3.gz. Build and install that and then we'll have a landmark
that I know works with the tar-1.15.1 builds.
The only thing that comes to mind quickly on that kernel if the version
string is correct, is the sparse file bug, that wasn't fixed till about
2.6.20-rc4. And it existed a very long time, as in clear back to 2.6.16
at least. I *think* that patch has been backported to at least as far
back as 2.6.17, even though the patch itself is pretty new.
The std test is to fire up a torrent client, and go get an install dvd,
all in one run of the client. When its done, check the sum. There is
about a 50% chance it will be bad, in which case stop, and restart the
client, which will cause it to recheck the file block by block, and
re-pull any bad blocks it finds. Repeat till the sum is good.
This bug only attacks where the whole incoming filesize is pre-allocated
on the drive as the torrent clients all do, and random writes into the
file are used to fill that space with data. No effect on normal
operations was ever detected that I heard about. That's why it took so
darned long to find it, nobody was getting bit often enough to make the
connection that it was a real bug. When it was declared a bug, Linus
held the 2.6.20-rc4 release up for nearly 2 weeks while everybody dropped
what they were doing and looked for that. ISTR it actually turned out to
be a typu.
>app-backup/amanda-2.5.1_p3-r2
>(unstable, but I know the ebuild-author ...)
I'm currently running the 2.5.1p3-20070312 snapshot, and AFAIAC its
stable. And I know the build author, me... And you've seen my script
often enough to have it memorized.
>I run this on one server with 3 clients, and incrementals happen all the
>time, so I think this works, at least here.
>
>Stefan
I really think we could be fighting with 2 separate problems. I'm trying
to get setup with quilt to do a bisect between 2.6.20.3 and 2.6.20.4-rc1,
only 31 patches, so maybe I can run the kernel bug down, I think it's in
one of those 31 patches. And as I said, this is not the first time such
a question has been asked about tar-1.16.
--
Cheers, Gene
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
When it's dark enough you can see the stars.
-- Ralph Waldo Emerson,
|
