Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Re: FreeBSD Port: lzma-4.48

0 views
Skip to first unread message

Andrew Pantyukhin

unread,
Jul 6, 2007, 2:01:01 PM7/6/07
to Eric Kingston, po...@freebsd.org
On 7/6/07, Eric Kingston <eri...@esreco.net> wrote:
> There is a bug in LZMA running on amd64. LZMA, an archiver port,
> segmentation faults on files > 1-5 GB in size. I've tested this on three
> different xeon dual multi-core processor servers. In each case, lzma core
> dumps in the exact same spot and the resulting file size for each is exactly
> the same. Files less than 1 GB seem to work ok. LZMA seems to work on any
> size file without a problem on the FreeBSD i386 platform. When I spoke with
> a friend of mine, he says that he compresses 160GB files daily without a
> problem, on his i386 systems.
>
> P.S. Here is my machine info..
>
> FreeBSD elrond.esreco.net 6.2-RELEASE-p5 FreeBSD 6.2-RELEASE-p5 #0: Sun Jun
> 17 13:37:57 MDT 2007
> eri...@elrond.esreco.net:/usr/obj/usr/src/sys/ELROND amd64

Thanks for your report! To achieve the best/quickest
possible result, could you also:
1. Try a previous version
2. File your bug report with the 7zip project
3. Get a gdb backtrace out of the coredump

I'm sorry but I don't have time to do it all right
now. I will look into it at the first opportunity,
though.

Thanks!

Peter Jeremy

unread,
Jul 6, 2007, 9:17:18 PM7/6/07
to Eric Kingston, po...@freebsd.org
On 2007-Jul-06 07:08:16 -0600, Eric Kingston <eri...@esreco.net> wrote:
>There is a bug in LZMA running on amd64. LZMA, an archiver port,
>segmentation faults on files > 1-5 GB in size. I've tested this on three
>different xeon dual multi-core processor servers. In each case, lzma core
>dumps in the exact same spot and the resulting file size for each is exactly
>the same.

I think this is data-dependent. I have tried compressing 4GB of arbitrary
data as well as >7GB of random data without problems. Are you able to
write a program to generate a sample dataset that displays the problem?

--
Peter Jeremy

0 new messages