This is the mail archive of the
cygwin-apps
mailing list for the Cygwin project.
Re: An all-new upset error message - tar related?
On Fri, Apr 14, 2006 at 05:17:21PM +0100, Dave Korn wrote:
>On 14 April 2006 17:14, Christopher Faylor wrote:
>
>> On Fri, Apr 14, 2006 at 05:08:52PM +0100, Dave Korn wrote:
>>> On 14 April 2006 16:59, Christopher Faylor wrote:
>>>
>>>> ----- Forwarded message from Cron Daemon <root@sourceware> -----
>>>>
>>>> Invalid header block at offset unknown at
>>>>
>>>
>/export/u0/sourceware/sourceware/libre/infra/bin/cygwin/Cygwin/Setup/Listing.p
>>> m
>>>> line 186 Invalid header block at offset unknown at
>>>>
>>>
>/export/u0/sourceware/sourceware/libre/infra/bin/cygwin/Cygwin/Setup/Listing.p
>>> m
>>>> line 186 Invalid header block at offset unknown at
>>>>
>>>
>/export/u0/sourceware/sourceware/libre/infra/bin/cygwin/Cygwin/Setup/Listing.p
>>> m
>>>> line 186 No data could be read from file at
>>>>
>>>
>/export/u0/sourceware/sourceware/libre/infra/bin/cygwin/Cygwin/Setup/Listing.p
>>> m
>>>> line 186 PROBLEMS WITH release/w32api/w32api-3.7-1-src.tar.bz2
>>>>
>>>> ----- End forwarded message -----
>>>>
>>>> I've been noticing this error message for a couple of days.
>>>> I wonder if it has something to do with the new version of tar?
>>>
>>> The message itself comes from bunzip2. If tar is somehow now piping stuff
>>> to or from it in textmode when using -j, .... ouch! The phrase "at offset
>>> unknown" suggests fseek-vs-textmode trouble as well.
>>>
>>>> So far, it has cropped up for monotone, coreutils, and, now, w32api.
>>>> Unpacking/repacking the files on sourceware.org seems to "fix" the
>>>> problem.
>>>
>>> ... which is a native LF-speaking linux box...
>>
>> And the actual error is coming from a native LF-speaking linux box. It's
>> hard to see how this could be a CRLF issue.
>
> I'm thinking of the pipe on the windows box where the tarball is packed up
>leaving CRLFs in the bunzip2 format which mess up the bunzip2 in the pipe on
>sourceware when it gets unpacked.
There is no windows box in the equation when I successfully unpack and
repack the archives on sourceware.
However, I just managed to duplicate the problem with a repacked tar.bz2 so
that should rule out a cygwin-tar problem. I'll investigate what changed on
sourceware.
One of the people with root access likes to "help out" by periodically
(and often silently) making changes on sourceware.org that make a lot of
sense to him -- even to parts of the system that I've historically
maintained. So, it's possible that something has been changed.
cgf