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

Re: svn commit r32706/32707/32730: Fix yet another issue #3067 bug

From: Paul Burba <ptburba_at_gmail.com>
Date: Tue, 26 Aug 2008 20:34:17 -0400

On Tue, Aug 26, 2008 at 7:15 PM, Julian Foad <julianfoad_at_btopenworld.com> wrote:
> Paul,
>
> I'm reviewing for back-port to 1.5.2.
>
>
>> r32706 | pburba | 2008-08-25 21:27:54 +0100 (Mon, 25 Aug 2008) | 7 lines
>> [[[
>> Add yet another issue #3067 merge test.
>>
>> * subversion/tests/cmdline/merge_tests.py
>> (merge_target_and_subtrees_need_nonintersecting_ranges): New.
>> (test_list): Add merge_target_and_subtrees_need_nonintersecting_ranges
>> and mark as XFail.
>> ]]]
>
>
>> +# Test for yet another variant of issue #3067.
>> +def merge_target_and_subtrees_need_nonintersecting_ranges(sbox):
>> + "target and subtrees need nonintersecting revs"
>> +
>> + sbox.build()
>> + wc_dir = sbox.wc_dir
>> +
>> + # Some paths we'll care about
>> + nu_path = os.path.join(wc_dir, "A", "D", "G", "nu")
>> + A_COPY_path = os.path.join(wc_dir, "A_COPY")
>> + nu_COPY_path = os.path.join(wc_dir, "A_COPY", "D", "G", "nu")
>> + omega_COPY_path = os.path.join(wc_dir, "A_COPY", "D", "H", "omega")
>> + beta_COPY_path = os.path.join(wc_dir, "A_COPY", "B", "E", "beta")
>> + beta_COPY_path = os.path.join(wc_dir, "A_COPY", "B", "E", "beta")
>> + rho_COPY_path = os.path.join(wc_dir, "A_COPY", "D", "G", "rho")
>> + omega_COPY_path = os.path.join(wc_dir, "A_COPY", "D", "H", "omega")
>> + psi_COPY_path = os.path.join(wc_dir, "A_COPY", "D", "H", "psi")
>
> omega_COPY_path and beta_COPY_path are each defined twice there. No
> harm, AFAIK, just odd.

Looks like a cut and paste error, fixed r32743.

>> +
>> + # Make a branch to merge to.
>> + wc_disk, wc_status = set_up_branch(sbox, False, 1)
>> +
>> + # Add file A/D/G/nu in r7.
>> + svntest.main.file_write(nu_path, "This is the file 'nu'.\n")
>> + svntest.actions.run_and_verify_svn(None, None, [], 'add', nu_path)
>> + expected_output = wc.State(wc_dir, {'A/D/G/nu' : Item(verb='Adding')})
>> + wc_status.add({'A/D/G/nu' : Item(status=' ', wc_rev=7)})
>> + svntest.actions.run_and_verify_commit(wc_dir, expected_output,
>> + wc_status, None, wc_dir)
>> +
>> + # Make a text mod to A/D/G/nu in r8.
>> + svntest.main.file_write(nu_path, "New content")
>> + expected_output = wc.State(wc_dir, {'A/D/G/nu' : Item(verb='Sending')})
>> + wc_status.tweak('A/D/G/nu', wc_rev=8)
>> + svntest.actions.run_and_verify_commit(wc_dir, expected_output,
>> + wc_status, None, wc_dir)
>> +
>> + # Do several merges to setup a situation where a subtree needs the merge
>> + # target and two of its subtrees all need non-intersecting ranges
>
> Are there a few superfluous words there? (s/a subtree needs //)

Fixed in r32743.

>> + # merged when doing a synch (a.k.a. cherry harvest) merge.
>> + #
>> + # 1) Merge -r0:7 from A to A_COPY.
>> + #
>> + # 2) Merge -c8 from A/D/G/nu to A_COPY/D/G/nu.
>> + #
>> + # 3) Merge -c-6 from A/D/H/omega to A_COPY/D/H/omega.
>> + #
>> + # Commit this group of merges as r9. Since we already test these type
>> + # of merges to death we don't use run_and_verify_merge() on these
>> + # intermediate merges.
>> + svntest.actions.run_and_verify_svn(
>> + None, expected_merge_output([[2,7]],
>> + ['U ' + beta_COPY_path + '\n',
>> + 'A ' + nu_COPY_path + '\n',
>> + 'U ' + rho_COPY_path + '\n',
>> + 'U ' + omega_COPY_path + '\n',
>> + 'U ' + psi_COPY_path + '\n']
>> + ),
>> + [], 'merge', '-r0:7', sbox.repo_url + '/A', A_COPY_path)
>> + svntest.actions.run_and_verify_svn(
>> + None, expected_merge_output([[8]], 'U ' + nu_COPY_path + '\n'),
>> + [], 'merge', '-c8', sbox.repo_url + '/A/D/G/nu', nu_COPY_path)
>> + svntest.actions.run_and_verify_svn(
>> + None, expected_merge_output([[-6]], 'G ' + omega_COPY_path + '\n'),
>> + [], 'merge', '-c-6', sbox.repo_url + '/A/D/H/omega', omega_COPY_path)
>> + expected_output = wc.State(wc_dir, {'A/D/G/rho' : Item(verb='Sending')})
>
> Redundant line: you overwrite expected_output with a different value 8
> lines later.

Fixed r32743.

>> + wc_status.add({'A_COPY/D/G/nu' : Item(status=' ', wc_rev=9)})
>> + wc_status.tweak('A_COPY',
>> + 'A_COPY/B/E/beta',
>> + 'A_COPY/D/G/rho',
>> + 'A_COPY/D/H/omega',
>> + 'A_COPY/D/H/psi',
>> + wc_rev=9)
>> + expected_output = wc.State(wc_dir, {
>> + 'A_COPY' : Item(verb='Sending'),
>> + 'A_COPY/B/E/beta' : Item(verb='Sending'),
>> + 'A_COPY/D/G/rho' : Item(verb='Sending'),
>> + 'A_COPY/D/G/nu' : Item(verb='Adding'),
>> + 'A_COPY/D/H/omega' : Item(verb='Sending'),
>> + 'A_COPY/D/H/psi' : Item(verb='Sending'),
>> + })
>> + svntest.actions.run_and_verify_commit(wc_dir, expected_output, wc_status,
>> + None, wc_dir)
>> +
>> + # Update the WC to allow full mergeinfo inheritance and elision.
>> + svntest.actions.run_and_verify_svn(None, ["At revision 9.\n"], [], 'up',
>> + wc_dir)
>> +
>> + # Merge all available revisions from A to A_COPY, the merge logic
>> + # should handle this situation (no "svn: Working copy path 'D/G/nu'
>> + # does not exist in repository" errors!) and after elision only
>> + # A_COPY should have explicit mergeinfo.
>> + short_A_COPY_path = shorten_path_kludge(A_COPY_path)
>> + expected_output = wc.State(short_A_COPY_path, {
>> + 'D/H/omega': Item(status='U '),
>> + })
>> + expected_status = wc.State(short_A_COPY_path, {
>> + '' : Item(status=' M', wc_rev=9),
>> + 'B' : Item(status=' ', wc_rev=9),
>> + 'mu' : Item(status=' ', wc_rev=9),
>> + 'B/E' : Item(status=' ', wc_rev=9),
>> + 'B/E/alpha' : Item(status=' ', wc_rev=9),
>> + 'B/E/beta' : Item(status=' ', wc_rev=9),
>> + 'B/lambda' : Item(status=' ', wc_rev=9),
>> + 'B/F' : Item(status=' ', wc_rev=9),
>> + 'C' : Item(status=' ', wc_rev=9),
>> + 'D' : Item(status=' ', wc_rev=9),
>> + 'D/G' : Item(status=' ', wc_rev=9),
>> + 'D/G/pi' : Item(status=' ', wc_rev=9),
>> + 'D/G/rho' : Item(status=' ', wc_rev=9),
>> + 'D/G/tau' : Item(status=' ', wc_rev=9),
>> + 'D/G/nu' : Item(status=' M', wc_rev=9),
>> + 'D/gamma' : Item(status=' ', wc_rev=9),
>> + 'D/H' : Item(status=' ', wc_rev=9),
>> + 'D/H/chi' : Item(status=' ', wc_rev=9),
>> + 'D/H/psi' : Item(status=' ', wc_rev=9),
>> + 'D/H/omega' : Item(status='MM', wc_rev=9),
>> + })
>> + expected_disk = wc.State('', {
>> + '' : Item(props={SVN_PROP_MERGEINFO : '/A:2-9'}),
>> + 'B' : Item(),
>> + 'mu' : Item("This is the file 'mu'.\n"),
>> + 'B/E' : Item(),
>> + 'B/E/alpha' : Item("This is the file 'alpha'.\n"),
>> + 'B/E/beta' : Item("New content"),
>> + 'B/lambda' : Item("This is the file 'lambda'.\n"),
>> + 'B/F' : Item(),
>> + 'C' : Item(),
>> + 'D' : Item(),
>> + 'D/G' : Item(),
>> + 'D/G/pi' : Item("This is the file 'pi'.\n"),
>> + 'D/G/rho' : Item("New content"),
>> + 'D/G/tau' : Item("This is the file 'tau'.\n"),
>> + 'D/G/nu' : Item("New content"),
>> + 'D/gamma' : Item("This is the file 'gamma'.\n"),
>> + 'D/H' : Item(),
>> + 'D/H/chi' : Item("This is the file 'chi'.\n"),
>> + 'D/H/psi' : Item("New content"),
>> + 'D/H/omega' : Item("New content"),
>> + })
>> + expected_skip = wc.State(short_A_COPY_path, { })
>> + saved_cwd = os.getcwd()
>> + os.chdir(svntest.main.work_dir)
>> + svntest.actions.run_and_verify_merge(short_A_COPY_path, None, None,
>> + sbox.repo_url + \
>> + '/A',
>> + expected_output,
>> + expected_disk,
>> + expected_status,
>> + expected_skip,
>> + None, None, None, None,
>> + None, 1)
>> + os.chdir(saved_cwd)
>> +
>
> OK, nothing in the test that prevents a successful review. +1 to r32706.

Sorry for the slop.

>> ------------------------------------------------------------------------
>> r32707 | pburba | 2008-08-25 22:01:28 +0100 (Mon, 25 Aug 2008) | 18 lines
>> [[[
>> Fix yet another issue #3067 bug.
>>
>> When driving the merge report editor and the merge target has subtrees that
>> need different ranges applied, there was a case we didn't handle: A subtree
>> may need a range merged that doesn't intersect with the range its nearest
>> parent requires *nor* that any other subtree requires. In some cases this
>> could cause us to describe path/revisions to the reporter that don't exist
>> resulting in the dreaded "svn: Working copy path 'blah' does not exist in
>> repository" errors.
>>
>> * subversion/libsvn_client/merge.c
>> (drive_merge_report_editor):
>
> It would be calming to see something written after the colon! Not
> necessary for successful review, though.

Tweaked the log message.

>> * subversion/tests/cmdline/merge_tests.py
>> (test_list): Remove XFail from
>> merge_target_and_subtrees_need_nonintersecting_ranges.
>> ]]]
>
>
>> /* If a subtree needs the same range applied as it's nearest parent
>> - with mergeinfo, then we don't need to describe the subtree
>> - separately. */
>> + with mergeinfo or neither the subtree nor its nearest parent need
>> + REVISION1:REVISION2 merged, then we don't need to describe the
>> + subtree separately. In the latter case this could break the
>> + editor if child->path didn't exist at REVISION2 and we attempt
>> + to describe it via a reporter set_path call. */
>> if (child->remaining_ranges->nelts)
>> {
>> range = APR_ARRAY_IDX(child->remaining_ranges, 0,
>> svn_merge_range_t *);
>> - if (parent->remaining_ranges->nelts)
>> + if ((!is_rollback && range->start > revision2)
>> + || (is_rollback && range->start < revision2))
>> + {
>> + /* Neither subtree nor parent need any
>> + part of REVISION1:REVISION2. */
>> + continue;
>> + }
>> + else if (parent->remaining_ranges->nelts)
>> {
>> svn_merge_range_t *parent_range =
>> APR_ARRAY_IDX(parent->remaining_ranges, 0,
>> @@ -2963,7 +2973,7 @@
>> APR_ARRAY_IDX(child->remaining_ranges, 0,
>> svn_merge_range_t *);
>> if (parent_range->start == child_range->start)
>> - continue; /* Same as parent. */
>> + continue; /* Subtree needs same range as parent. */
>> }
>> }
>> else /* child->remaining_ranges->nelts == 0*/
>
> I'm still studying the implementation.

Thanks for checking this out Julian, let me know if there are more
questions (or just ping me in IRC in the morning).

Paul

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-08-27 02:39:25 CEST

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.