I hate to be a heartless bastard, but I'm going to have to pass on
trying to reproduce this; it just seems like there's a high risk of
being unable to. If you can make a reproducible case (ideally, a
script we can run to make it happen), we can look at the bug.
And of course, I speak only for myself when it comes to allocation of
time. Someone else may decide to try reproducing this, and I hope
they succeed if so.
Clem Taylor <firstname.lastname@example.org> writes:
> On 28 Nov 2005 11:14:08 -0600, email@example.com <firstname.lastname@example.org> wrote:
> > I can't tell if you have found a bug here or not yet. Would it be
> > possible to get a transcript? (Verbs "updated this tree" and "backed
> > out this change" are okay in conversation, but for a reproduction
> > recipe we need more precision -- what exact commands did you use?)
> I can't easily provide a transcript for the vendor tree. Basically, I
> removed all the files in a working copy, leaving the directories and
> .svn stuff alone. Next, I copied the new version (from a git tree) and
> ran a script that deletes missing files and adds new files. I checked
> in the result at tagged it 2.6.14. I've done this many times without
> any problems. However in this case it turned out that what I thought
> was 2.6.14 was actually 2.6.14 plus some badly broken stuff. So, the
> "backed out this change" step just repeated the above, copying a
> different git view. I deleted the old 2.6.14 tag and created a new
> 2.6.14 tag. The problem is that the 2.6.14+ version had a huge number
> of changes and a number of files that were deleted. When I checked in
> the proper 2.6.14 code, some of the ancestry was lost.
> So next I wanted to merge my 2.6.13 based devel tree with the new 2.6.14 tree:
> # check out clean working copy of the trunk
> $ svn co svn+ssh://.../trunk/np/linux
> $ svn status -u
> ... no error ...
> # merge previous linux tag with new linux tag from vendor tree
> $ svn merge svn+ssh://.../vendor/linux-mips/2.6.13 \
> svn+ssh://.../vendor/linux-mips/2.6.14 .
> ... 645 deletes, 733 adds, 4227 updates, 1 'Skipped missing target' ...
> ... NOTE: The merge doesn't finish, it fails with:
> svn: URL 'svn+ssh://.../vendor/linux-mips/2.6.14/drivers/acpi/ospm'
> doesn't match existing
> URL 'svn+ssh://.../trunk/np/linux/drivers/acpi/ospm' in 'drivers/acpi/ospm'
> # at this point the working copy is broken:
> $ svn status -u
> svn: Working copy path 'include/asm-ppc64/hw_irq.h' does not exist in repository
> $ svn status include/asm-ppc64/hw_irq.h
> R + include/asm-ppc64/hw_irq.h
> $ svn info include/asm-ppc64/hw_irq.h
> Path: include/asm-ppc64/hw_irq.h
> Name: hw_irq.h
> URL: svn+ssh://.../trunk/np/linux/include/asm-ppc64/hw_irq.h
> Revision: 0
> Node Kind: file
> Schedule: replace
> Copied From URL:
> Copied From Rev: 1089
> Last Changed Author: ctaylor
> Last Changed Rev: 4
> Last Changed Date: 2005-03-18 01:14:14 -0500 (Fri, 18 Mar 2005)
> Text Last Updated: 2005-11-28 14:06:09 -0500 (Mon, 28 Nov 2005)
> Properties Last Updated: 2005-11-28 14:06:09 -0500 (Mon, 28 Nov 2005)
> Checksum: eb2908eef2a030257f9ec40b9543d23d
> One interesting thing the merge had a number of lines like:
> D include/asm-ppc64/hw_irq.h
> A include/asm-ppc64/hw_irq.h
> So the file was deleted (2.6.14+) and then added back (2.6.14). This
> seems to be the root of the problem.
> I tried doing the merge with '--ignore-ancestry', but this didn't work
> either, it failed with:
> Skipped missing target: 'kernel/fork.c'
> ... many more skipped missing targets ...
> Skipped missing target: 'include/asm-ia64/socket.h'
> Skipped 'include/asm-ia64/hdreg.h'
> Skipped missing target: 'include/asm-ia64'
> svn: Working copy 'include' is missing or not locked
To unsubscribe, e-mail: email@example.com
For additional commands, e-mail: firstname.lastname@example.org
Received on Tue Nov 29 00:09:54 2005