Re: svn commit: r942170 - in /subversion/trunk/subversion/libsvn_wc: props.c props.h status.c

["Hyrum K. Wright" <hy...@mail.utexas.edu>]

On Fri, May 7, 2010 at 1:34 PM, <gs...@apache.org> wrote:

> Author: gstein
> Date: Fri May  7 18:34:12 2010
> New Revision: 942170
>
> URL: http://svn.apache.org/viewvc?rev=942170&view=rev
> Log:
> Remove a couple internal props functions.
>
> * subversion/libsvn_wc/props.h:
>  (svn_wc__has_props): removed
>  (svn_wc__wcprop_set): declaration removed.
>
> * subversion/libsvn_wc/props.c:
>  (svn_wc__wcprop_set): renamed to ...
>  (wcprop_set): ... this. made static. added comment about the
>    transaction-ness of this function.
>  (svn_wc__internal_propset): track rename to wcprop_set
>  (svn_wc__has_props): removed
>
> * subversion/libsvn_wc/status.c:
>  (assemble_status): alter call to __has_props by fetching the pristine
>    and current set of properties for the node. leave a note that we
>    cannot (yet) rely on these for the __props_modified call.
>
> Modified:
>    subversion/trunk/subversion/libsvn_wc/props.c
>    subversion/trunk/subversion/libsvn_wc/props.h
>    subversion/trunk/subversion/libsvn_wc/status.c

...


I think this is causing the JavaHL test failure, namely that status now
fails on a missing, schedule-delete directory.  Could this be a connection?

-Hyrum

Re: svn commit: r942170 - in /subversion/trunk/subversion/libsvn_wc: props.c props.h status.c

[Greg Stein <gs...@gmail.com>]

On Tue, May 11, 2010 at 14:17, Greg Stein <gs...@gmail.com> wrote:
> On Tue, May 11, 2010 at 12:39, Philip Martin <ph...@wandisco.com> wrote:
>...
>> It appears that the main testsuite doesn't check status on missing,
>> deleted directories.  942170 is when a simple 'svn st' started
>> failing, 'svn st -u' has been failing for much longer.  There is some

The key piece was an svn_wc__adm_missing() check before recursing on a
subdirectory. I removed that in r37361 a year ago, and put an
svn_wc__get_entry() call in its place. That would have error'd if it
could not retrieve the "real" data. (tho the error check was wrong...
no biggy)

Last August, r38991 replaced all the entry logic with some other wc-ng
gunk, and the check was lost entirely.

So today, the crawler attempts to recurse "into" an obstructing file. Whoops.

This should be a pretty simple fix...

Cheers,
-g

Re: svn commit: r942170 - in /subversion/trunk/subversion/libsvn_wc: props.c props.h status.c

[Greg Stein <gs...@gmail.com>]

On Tue, May 11, 2010 at 12:39, Philip Martin <ph...@wandisco.com> wrote:
> "Hyrum K. Wright" <hy...@mail.utexas.edu> writes:
>
>> I think this is causing the JavaHL test failure, namely that status now
>> fails on a missing, schedule-delete directory.  Could this be a connection?
>
> Well check-javahl with 942169 passes and with 942170 it fails :)

Yup.

> It appears that the main testsuite doesn't check status on missing,
> deleted directories.  942170 is when a simple 'svn st' started
> failing, 'svn st -u' has been failing for much longer.  There is some
> code in basic_tests.py:basic_revert:828 that would check this but it
> is commented out.

I'll write a specific test for obstructions and properties, and then
get this bug fixed up.

Thanks,
-g

Re: svn commit: r942170 - in /subversion/trunk/subversion/libsvn_wc: props.c props.h status.c

[Philip Martin <ph...@wandisco.com>]

"Hyrum K. Wright" <hy...@mail.utexas.edu> writes:

> I think this is causing the JavaHL test failure, namely that status now
> fails on a missing, schedule-delete directory.  Could this be a connection?

Well check-javahl with 942169 passes and with 942170 it fails :)

It appears that the main testsuite doesn't check status on missing,
deleted directories.  942170 is when a simple 'svn st' started
failing, 'svn st -u' has been failing for much longer.  There is some
code in basic_tests.py:basic_revert:828 that would check this but it
is commented out.

-- 
Philip