You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Erik Huelsmann <eh...@gmail.com> on 2006/09/06 11:58:10 UTC

Re: svn: Checksum mismatch

On 9/5/06, Jens Seidel <je...@users.sf.net> wrote:
> Hi,
>
> I have good news. I was able to create a small example which demonstrates
> the checksum error in Version 1.4.0 (Release Candidate 5).

> I attached a script which shows this error, here is the output:
>
> + mkdir /tmp/svn.test
> + cd /tmp/svn.test
> + svnadmin create repos
> + R=file:///tmp/svn.test/repos
> + svn co file:///tmp/svn.test/repos wc
> Checked out revision 0.
> + cd wc
> + mkdir trunk
> + mkdir trunk/project
> + echo 'file 1'
> + svn add trunk
> A         trunk
> A         trunk/project
> A         trunk/project/file1.old
> + svn ci '-mInitial version'
> Adding         trunk
> Adding         trunk/project
> Adding         trunk/project/file1.old
> Transmitting file data .
> Committed revision 1.
> + svn move '-mCreated file1' file:///tmp/svn.test/repos/trunk/project/file1.old file:///tmp/svn.test/repos/trunk/project/file1
>
> Committed revision 2.
> + svn copy '-mCreated branch' file:///tmp/svn.test/repos/trunk file:///tmp/svn.test/repos/branch
>
> Committed revision 3.
> + svn move '-mMoved trunk to trunk.old' file:///tmp/svn.test/repos/trunk file:///tmp/svn.test/repos/trunk.old
>
> Committed revision 4.
> + svn mkdir '-mCreated new trunk' file:///tmp/svn.test/repos/trunk
>
> Committed revision 5.
> + svn copy '-mRestore old trunk' -r 1 file:///tmp/svn.test/repos/trunk/project file:///tmp/svn.test/repos/trunk/src
>
> Committed revision 6.
> + svn up
> D    trunk
> A    trunk
> A    trunk/src
> A    trunk/src/file1.old
> A    trunk.old
> A    trunk.old/project
> A    trunk.old/project/file1
> A    branch
> A    branch/project
> A    branch/project/file1
> Updated to revision 6.
> + echo 'file 1 (new)'
> + svn add trunk/src/file1
> A         trunk/src/file1
> + svn ci '-mAdded file1'
> Adding         trunk/src/file1
> Transmitting file data .
> Committed revision 7.
> + echo 'file 1 (new)'
> + svn ci '-mManual sync against trunk' branch/project/file1
> Sending        branch/project/file1
> Transmitting file data .
> Committed revision 8.
> + oldRev=7
> + cd branch
> + svn up -r 7
> U    project/file1
> Updated to revision 7.
> + cd project
> + svn merge file:///tmp/svn.test/repos/trunk/project@3 file:///tmp/svn.test/repos/trunk/src@7
> D    file1
> A    file1
> A    file1.old
> + svn update
> svn: Checksum mismatch for '.svn/text-base/file1.svn-base'; expected: 'e243bb39c844b3543a7726576c869caf', actual: '0ad777f8c28bdfc866607b87f653b5e8'


I have been looking at svn_wc_crawl_revisions2 which drives the
revision reporter for updates. I must assume the working copy isn't
marked as incomplete, so that the crawler reports the entire working
copy. It contains this segment:

           if (report_everything)
             {
               /* Report the file unconditionally, one way or another. */
               if (strcmp(current_entry->url, this_url) != 0)
                 SVN_ERR(reporter->link_path(report_baton, this_path,
                                             current_entry->url,
                                             current_entry->revision,
                                             FALSE, current_entry->lock_token,
                                             iterpool));
               else
                 SVN_ERR(reporter->set_path(report_baton, this_path,
                                            current_entry->revision,
                                            FALSE, current_entry->lock_token,
                                            iterpool));
             }

But, in case of replaced paths, I think that the driver should call
reporter->delete_path() for replace-no-history and
reporter->link_path() for replace-with-history schedules.

There are some oddities wrt replaced paths a bit further down too (why
don't we ->link_path() for replace-with-history scheduled files
everywhere? Using set_path implies we want to receive updates from the
file in the current location...)

Opinions?

bye,


Erik.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn: Checksum mismatch

Posted by Erik Huelsmann <eh...@gmail.com>.
On 9/28/06, Jens Seidel <je...@users.sf.net> wrote:
> Hi Erik,
>
> what's the status of this bug? Is it already fixed in trunk?
> If not I suggest you open a bug report to ensure we do not forget about
> it ...
>
> On Wed, Sep 06, 2006 at 09:47:15AM -0400, Garrett Rooney wrote:
> > On 9/6/06, Erik Huelsmann <eh...@gmail.com> wrote:
> > >On 9/5/06, Jens Seidel <je...@users.sf.net> wrote:
> > >> I have good news. I was able to create a small example which demonstrates
> > >> the checksum error in Version 1.4.0 (Release Candidate 5).
> > >
> > >I have been looking at svn_wc_crawl_revisions2 which drives the
>
> [snip]
>
> > >But, in case of replaced paths, I think that the driver should call
> > >reporter->delete_path() for replace-no-history and
> > >reporter->link_path() for replace-with-history schedules.
> >
> > That seems reasonable to me, although I don't know nearly enough about
> > how this stuff works, so take that with a grain of salt...
> >
> > >There are some oddities wrt replaced paths a bit further down too (why
> > >don't we ->link_path() for replace-with-history scheduled files
> > >everywhere? Using set_path implies we want to receive updates from the
> > >file in the current location...)
> >
> > It does seem like replacements should be handled specially there...

I committed a fix to trunk in r21720.

> Jens

HTH,

Erik.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn: Checksum mismatch

Posted by Jens Seidel <je...@users.sourceforge.net>.
Hi Erik, 

what's the status of this bug? Is it already fixed in trunk?
If not I suggest you open a bug report to ensure we do not forget about
it ...

On Wed, Sep 06, 2006 at 09:47:15AM -0400, Garrett Rooney wrote:
> On 9/6/06, Erik Huelsmann <eh...@gmail.com> wrote:
> >On 9/5/06, Jens Seidel <je...@users.sf.net> wrote:
> >> I have good news. I was able to create a small example which demonstrates
> >> the checksum error in Version 1.4.0 (Release Candidate 5).
> >
> >I have been looking at svn_wc_crawl_revisions2 which drives the

[snip]

> >But, in case of replaced paths, I think that the driver should call
> >reporter->delete_path() for replace-no-history and
> >reporter->link_path() for replace-with-history schedules.
> 
> That seems reasonable to me, although I don't know nearly enough about
> how this stuff works, so take that with a grain of salt...
> 
> >There are some oddities wrt replaced paths a bit further down too (why
> >don't we ->link_path() for replace-with-history scheduled files
> >everywhere? Using set_path implies we want to receive updates from the
> >file in the current location...)
> 
> It does seem like replacements should be handled specially there...

Jens

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn: Checksum mismatch

Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On 9/6/06, Erik Huelsmann <eh...@gmail.com> wrote:
> On 9/5/06, Jens Seidel <je...@users.sf.net> wrote:
> > Hi,
> >
> > I have good news. I was able to create a small example which demonstrates
> > the checksum error in Version 1.4.0 (Release Candidate 5).
>
> > I attached a script which shows this error, here is the output:
> >
> > + mkdir /tmp/svn.test
> > + cd /tmp/svn.test
> > + svnadmin create repos
> > + R=file:///tmp/svn.test/repos
> > + svn co file:///tmp/svn.test/repos wc
> > Checked out revision 0.
> > + cd wc
> > + mkdir trunk
> > + mkdir trunk/project
> > + echo 'file 1'
> > + svn add trunk
> > A         trunk
> > A         trunk/project
> > A         trunk/project/file1.old
> > + svn ci '-mInitial version'
> > Adding         trunk
> > Adding         trunk/project
> > Adding         trunk/project/file1.old
> > Transmitting file data .
> > Committed revision 1.
> > + svn move '-mCreated file1' file:///tmp/svn.test/repos/trunk/project/file1.old file:///tmp/svn.test/repos/trunk/project/file1
> >
> > Committed revision 2.
> > + svn copy '-mCreated branch' file:///tmp/svn.test/repos/trunk file:///tmp/svn.test/repos/branch
> >
> > Committed revision 3.
> > + svn move '-mMoved trunk to trunk.old' file:///tmp/svn.test/repos/trunk file:///tmp/svn.test/repos/trunk.old
> >
> > Committed revision 4.
> > + svn mkdir '-mCreated new trunk' file:///tmp/svn.test/repos/trunk
> >
> > Committed revision 5.
> > + svn copy '-mRestore old trunk' -r 1 file:///tmp/svn.test/repos/trunk/project file:///tmp/svn.test/repos/trunk/src
> >
> > Committed revision 6.
> > + svn up
> > D    trunk
> > A    trunk
> > A    trunk/src
> > A    trunk/src/file1.old
> > A    trunk.old
> > A    trunk.old/project
> > A    trunk.old/project/file1
> > A    branch
> > A    branch/project
> > A    branch/project/file1
> > Updated to revision 6.
> > + echo 'file 1 (new)'
> > + svn add trunk/src/file1
> > A         trunk/src/file1
> > + svn ci '-mAdded file1'
> > Adding         trunk/src/file1
> > Transmitting file data .
> > Committed revision 7.
> > + echo 'file 1 (new)'
> > + svn ci '-mManual sync against trunk' branch/project/file1
> > Sending        branch/project/file1
> > Transmitting file data .
> > Committed revision 8.
> > + oldRev=7
> > + cd branch
> > + svn up -r 7
> > U    project/file1
> > Updated to revision 7.
> > + cd project
> > + svn merge file:///tmp/svn.test/repos/trunk/project@3 file:///tmp/svn.test/repos/trunk/src@7
> > D    file1
> > A    file1
> > A    file1.old
> > + svn update
> > svn: Checksum mismatch for '.svn/text-base/file1.svn-base'; expected: 'e243bb39c844b3543a7726576c869caf', actual: '0ad777f8c28bdfc866607b87f653b5e8'
>
>
> I have been looking at svn_wc_crawl_revisions2 which drives the
> revision reporter for updates. I must assume the working copy isn't
> marked as incomplete, so that the crawler reports the entire working
> copy. It contains this segment:
>
>            if (report_everything)
>              {
>                /* Report the file unconditionally, one way or another. */
>                if (strcmp(current_entry->url, this_url) != 0)
>                  SVN_ERR(reporter->link_path(report_baton, this_path,
>                                              current_entry->url,
>                                              current_entry->revision,
>                                              FALSE, current_entry->lock_token,
>                                              iterpool));
>                else
>                  SVN_ERR(reporter->set_path(report_baton, this_path,
>                                             current_entry->revision,
>                                             FALSE, current_entry->lock_token,
>                                             iterpool));
>              }
>
> But, in case of replaced paths, I think that the driver should call
> reporter->delete_path() for replace-no-history and
> reporter->link_path() for replace-with-history schedules.

That seems reasonable to me, although I don't know nearly enough about
how this stuff works, so take that with a grain of salt...

> There are some oddities wrt replaced paths a bit further down too (why
> don't we ->link_path() for replace-with-history scheduled files
> everywhere? Using set_path implies we want to receive updates from the
> file in the current location...)

It does seem like replacements should be handled specially there...

-garrett

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org