[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: svn commit: r921445 - in /subversion/trunk/subversion: include/private/svn_wc_private.h include/svn_wc.h libsvn_client/commit.c libsvn_wc/entries.c

From: Joe Swatosh <joe.swatosh_at_gmail.com>
Date: Thu, 11 Mar 2010 19:22:59 -0800

On Wed, Mar 10, 2010 at 10:23 PM, Greg Stein <gstein_at_gmail.com> wrote:
> On Thu, Mar 11, 2010 at 01:06, Joe Swatosh <joe.swatosh_at_gmail.com> wrote:
>> On Wed, Mar 10, 2010 at 9:15 PM, Greg Stein <gstein_at_gmail.com> wrote:
>>> On Wed, Mar 10, 2010 at 21:23, Joe Swatosh <joe.swatosh_at_gmail.com> wrote:
>>>> Hi Philip
>>>> On Wed, Mar 10, 2010 at 8:47 AM,  <philip_at_apache.org> wrote:
>>>>> Author: philip
>>>>> Date: Wed Mar 10 16:47:27 2010
>>>>> New Revision: 921445
>>>>> URL: http://svn.apache.org/viewvc?rev=921445&view=rev
>>>>> Log:
>>>>> Remove some access batons from post-commit processing.
>>>> Since this patch the Ruby bindings test has been failing on the Ubuntu
>>>> build bot (and locally for me).  I haven't looked at the patch closely
>>>> so I'm not sure what's up, but the relevant part of the test is:
>>>>      Svn::Wc::AdmAccess.open(nil, @wc_path, true, 5) do |access|
>>>>        assert_raises(Svn::Error::WcPathFound) do
>>>>          access.mark_missing_deleted(path1)
>>>>        end
>>>>        FileUtils.rm(path1)
>>>>        access.mark_missing_deleted(path1)  ### This line raises
>>>> Svn::Wc::WcPathFound ###
>>>>        access.maybe_set_repos_root(path2, @repos_uri)
>>>>      end
>>>> Judging by the name of the function, it appears to me that the test is
>>>> correct....
>>>> I double checked, the rm is removing the file before the call.  Could
>>>> you check this please, or suggest how the test should be modified?
>>> The problem is that svn_wc__temp_mark_missing_not_present() now does a
>>> call to read_info(), whereas before it simply checked the disk. The
>>> function is verifying more state before progressing. In particular,
>>> you must have a directory that is scheduled for deletion before
>>> calling that function. Does your test properly set that up?
>>> Cheers,
>>> -g
>> Here is the current doc string for svn_wc_mark_missing_deleted:
>> /** Mark missing @a path as 'deleted' in its @a parent's list of entries.
>>  *
>>  * Return #SVN_ERR_WC_PATH_FOUND if @a path isn't actually missing.
>>  *
>>  * @deprecated Provided for backward compatibility with the 1.6 API.
>>  */
>> I want to make sure I'm understanding correctly.  It sounds like I need to
>> svn del 'path's parent' before calling svn_wc_mark_missing_deleted on 'path.'
>> Did I follow?  If so, then no the test doesn't set that up.
> Sorry I wasn't clear. The function is only useful for situations like:

Not sure it was you who wasn't clear :-)

> $ svn rm path/some-subdir
> $ rm -r path/some-subdir
> ... and then calling the function on path/some-subdir. It will record
> (in path/.svn/) that some-subdir is "not-present" in the working copy.
> The function is used as part of the commit code. Post-commit, we are
> allowed to mark that subdir as "not-present" in the new revision of
> path/... since we know it was deleted as part of the commit. We don't
> mind that the subdir is missing because that recording is in the stub
> located in path/.svn/
> That said, the function is bogus because this kind of recording should
> happen during commit processing naturally. And it should bump the
> revision on that directory, rather than just alter the presence value
> stashed away in the stub. We should be able to eliminate the function
> today, and fold it into commit processing (and then it *may* further
> simplify during the single-db move).

Thanks for the explanation. I'm guessing your advice would be to dump
the broken test of the deprecated function (rather than fix the test)?

Received on 2010-03-12 04:23:31 CET

This is an archived mail posted to the Subversion Dev mailing list.