STATUS - subversion - Git at Google

       * * * * * * * * * * * * * * * * * * * * * * * * * * * *
       *                                                     *
       *  THIS RELEASE STREAM IS OPEN TO BUG FIXES.          *
       *                                                     *
       * * * * * * * * * * * * * * * * * * * * * * * * * * * *

 This file tracks the status of releases in the 1.8.x line.

 See http://subversion.apache.org/docs/community-guide/releasing.html#release-stabilization
 for details on how release lines and voting work, what kinds of bugs can
 delay a release, etc.

 Status of 1.8.9:

 Candidate changes:
 ==================

  * r1469519, r1469520, r1481848, r1481870
    Silence compiler warnings, tighten scopes and improve code.
    Justification:
      Clean builds make it more likely to notice real compiler warnings.
    Branch: ^/subversion/branches/1.8.x-r1469519-and-friends
    Votes:
      +1: blair, brane
      -0: philip (see dev thread: false alarm, trades one warning for another)

  * r1501293, r1501338, r1501360, r1501533
    Fix issue #4390, "parallel nested checkout not possible with wc-ng"
    Justification:
      Regression from 1.6. User complained. See issue for details.
    Branch:
      ^/subversion/branches/1.8.x-issue4390
    Depends:
      Nomination conflicts with the branch and merges cleanly without it.
      Branch should probably be removed from the nomination (it was only needed
      before r1503552).
    Votes:
      +1: stsp
      -0: ivan, brane (not regression from 1.7.x,
                       destructive change in libsvn_wc core)
      -0: rhuijben (temporarily remove my +1 until I complete investigating
                    a related problem. It looks like this fix doesn't work for
                    the user scenario it was intended for)

  * r1531014, r1532316, r1536350, r1536383, r1536464, r1536488, r1567204, r1567228
    Fix working copy database inconsistency on certain move scenarios,
    including those described in issue #4437 "Deleting a moved file can maintain a
    half move"
    Justification:
      The working copy database should be consistent on any normal operation.
    Branch:
      ^/subversion/branches/1.8.x-issue4437
    Votes:
      +1: rhuijben
      +1: (before branch, without r1567204, r1567228) philip

  * r1518942, r1536914, r1536931
    Fix commit of a specific mixed revision copy with non copy descendants that
    shadow not present nodes. This problem manifests itself as a copy of a tree
    that can't be committed because it is reported out of date.
    Justification:
      'svn cp TREE NEW-TREE; svn ci NEW-TREE' should never report out of date
      on some random descendant.
    Branch: ^/subversion/branches/1.8.x-r1536931
    Votes:
      +1: rhuijben

  * r1536854
    Make 'svnadmin verify' detect inconsistencies that will prevent loading
    dump files.
    Justification:
      Some users rely on dump files as a means of repository backup.  Without
      this patch, there is no way except of 'svnadmin load' to know that these
      dump files will load at all.  With this patch, a successful verify run
      should guarantee loadable dump files.
    Branch: ^/subversion/branches/1.8.x-r1536854
    Votes:
      +1: stefan2

  * r1538812, r1538925, r1538928, r1538939
    Properly remove moved_to information when the node in its original location
    is removed.
    Justification:
      The NODES table in the working copy database should always be consistent.
    Notes:
      rhuijben: The test for this problem might accidentally depend on other
                move fixes on trunk.
    Votes:
      +1: rhuijben, philip

  * r1542610, r1564576, r1568180
    Add missing description to build.conf, to properly generate a version
    resource (including icon) for this executable on Windows.
    Notes:
      r1564576 breaks compilation on Visual Studio 2010, 2012 and 2013 of
      this project. r1568180 fixes this problem.
    Votes:
      +1: rhuijben, breser
      +1: (without r1564567, r1568180): stsp
      -0: brane (horrible grammar; suggest translation to English)

  * r1544597, r1544600, r1544688, r1544690, r1545111, r1545122
    Fix a specific case where the api behind status would report a different
    move status for A depending on whether the user called 'svn status A' or
    'svn status wc' and then looked at 'wc/A'.
    Justification:
      Api users such as AnkhSVN rely on the result of svn status to match and
      this breaks their assumptions.
      This change has the nice side effect that it resolves the most common
      user triggerable assertion in AnkhSVN, which is really caused by a
      broken move. (See other nomination for a fix). But even then it shouldn't
      matter how status is called: on the node or an ancestor.
      (It also improves status performance a tiny bit, by reducing the number
       of sqlite transactions used)
    Notes:
      r1564587 fixes a function reference on the branch in some code that
      is disabled by the macro processor on Windows.
    Branch:
      ^/subversion/branches/1.8.x-r1544597
    Votes:
      +1: rhuijben
      -0 (before branch): philip (Doesn't build, needs a branch due to r1539279)

  * r1555491
    Fix svn_fs_close_root for FSFS transaction roots.
    Justification:
      Without this patch, cache-related pool cleanups may be run after the
      svn_fs_t object has been destroyed.  I.e. even with explicit calls to
      svn_fs_close_root, the user may see a segfault.
    Branch:
      ^/subversion/branches/1.8.x-r1555491
    Votes:
      +1: stefan2
      -0: philip, rhuijben (Seems to be a different name for the same pool?)

  * r1547454
    In FSFS, omit config file when creating pre-1.5 format repositories.
    Justification:
      Old SVN versions won't preserve this file in hotcopy.  Later versions
      might also behave inconsistently in the presence of this file if their
      feature availability check is incomplete.
    Votes:
      +1: stefan2
      -0: rhuijben (Might require r1561703 to avoid breaking tests)

  * r1561426
    In FSFS, skip the generation of uniquifiers if rep-shared is no supported.
    Branch: ^/subversion/branches/1.8.x-r1561426
    Justification:
      This is not known to be a bug, atm, as we omit these when we write the
      final revisions.  But we rely on those uniquifiers to be NULL in old
      repo formats when we compare reps.  Them being non-NULL in a txn might
      cause problems with future changes.
    Votes:
      +1: stefan2

  * r1567996, r1568009, r1568062, r1568063
    Fix a caching inefficiency in multi-threaded servers.
    Justification:
      .
    Notes:
      Due to going back and forth with different approaches to solve this
      issue, the best way to review this
    Branch:
      ^/subversion/branches/1.8.x-r1567996
    Votes:
      +1: stefan2
      -0: rhuijben (incomplete (misses cache->total_reads accesses) and
                    overkill. The whole thing can be implemented with simple
                    atomic operations while mutexes may be a limited resource)

  * r1571747, r1571787, r1571795
    Resolve the 'commit_util.c' (repos_root_url && repos_relpath) assertion.
    Notes:
      This assertion is raised when specifically committing a deleted
      descendant, like via a checkbox list in a GUI client.
    Justification:
      Often reported assertion by users. Upto now very hard to reproduce
      for these users. Probably because they didn's see that the node they
      deleted wasn't added to the repository yet.
    Votes:
      +1: rhuijben

  * r1572102, r1572200
    Fix a failed wc upgrade from 1.7 with exclusive wc locking.
    Notes:
      1.8 doesn't support running the regression test with exclusive locking
      so the regression test always passes.
    Justification:
      Reported by a user, upgrade is expected to work.
    Votes:
      +1: philip, brane

  * r1573088
    Add Mavericks to our sysinfo output.
    Justification:
      It's obvious, keeps us consistent with previous version behavior.
    Votes:
      +1: breser, brane

  * r1575270, r1575284
    Fix inconsistent output in svn_client_proplist4 when using absolute paths.
    Notes:
      r1575284 is the test that exposes this bug.
    Justification:
      It's a regression in 1.8.8 compared to 1.8.5.
    Votes:
       +1: brane, rhuijben

  * r1574868
    Don't let invalid mergeinfo stop 'svnadmin dump' from producing a dump.
      Part of issue #4476 "Mergeinfo containing r0 makes svnsync and svnadmin
      dump fail".
    Justification:
      A formatting error in what is essentially client-side metadata, albeit
      with repository-side interpretation as well, should never prevent
      dumping the repository. This issue only occurs when svnadmin is trying
      to warn us of mergeinfo revisions that might be unexpected, so it is
      not essential.
    Notes:
      There is a work-around, at least in theory: a dump starting from
      revision 0 or 1 does not suffer from this problem.
    Branch:
      ^/subversion/branches/1.8.x-r1574868
    Votes:
      +1: julianfoad

  * r1575525, r1575915
    Resolve issue #4427: File externals can't be exported.
    Justification:
      File externals 'are just files', so why should they be discriminated
      by export?
    Notes:
      r1575525 resolves the most important issues, but in some cases still
      doesn't allow exporting file externals. r1575915 simplifies the fix and
      resolves a few remaining issues.
    Votes:
      +1: rhuijben

  * r1577151
    Make 'svn merge' honor the 'preserved-conflict-file-exts' setting.
    Justification:
      While really a behavior change that doesn't belong in a patch release
      I'm guessing nobody knew that we didn't implement this flag in this
      scenario.
    Votes:
      +1: rhuijben

  * r1577294
    Make interactive conflict resolver behave as designed.
    Justification:
      This was intended to be the behavior form the start.  A typo in a
      refactoring broke it.  This fix helps users avoid accidentally choosing
      'mark resolved' and committing conflict markers.
    Votes:
      +1: breser

  * r1577812, r1577813
    Resolve a segfault when 'svn merge --force' merges a directory delete
    Justification:
      Even though this merge flag is not recommended, it shouldn't cause a
      segfault.
    Notes:
      r1577812 contains the actual fix and a regression test. r1577813 is
      a minor improvement. A branch is necessary because a comment right above
      the fix contained a typo that was fixed on trunk in a huge commit of many
      typo fixes.
    Branch:
      ^/subversion/branches/1.8.x-r1577812
    Votes:
      +1: rhuijben

  * r1578273, r1578311, r1578326
    Make svn_ra_get_locks() and svn_ra_get_lock() report not locked nodes
    with a NULL svn_lock_t *, as documented.
    Justification:
      Many clients use the existance of an svn_lock_t on a node via info/status
      as a boolean to note that there is a lock. But because we didn't properly
      check the result we reported mostly empty svn_lock_t instances in more
      cases. Even on directories!
    Notes:
      A branch is needed, because the serf lock handling code was refactored and
      split to multiple files on trunk.
    Branch:
      ^/subversion/branches/1.8.x-r1578311
    Votes:
      +1: rhuijben

  * r1578670, r1578820
    Fix the order of node record headers written by svndumpfilter.
    Justification:
      Makes output match the rules documented in notes/dump-load-format.txt.
    Votes:
      +1: philip, rhuijben

 Veto-blocked changes:
 =====================

  * r1524145
    Make the 'm' conflict prompt option run external merge tool if configured.
    Justification:
      Usability fix. The current behaviour ('m' -> internal merge tool,
      'l' -> external merge tool) confuses users who try 'm' to run an
      external merge tool. See issue #4426.
    Votes:
      +1: stsp
      -1: breser (see my email to dev@, <5282FE15.40505@reser.org>)

  * r1560723, r1561419, r1561427, r1561428
    Fix the handling of pre-1.4 FSFS repositories.
    Branch: ^/subversion/branches/1.8.x-r1560723-and-friends
    Justification:
      Without this, old SVN versions won't be able to e.g. hotcopy those
      repos once 1.8.x made a commit in them.  Also, fix our hotcopy code
      and make it cope with repositories broken by previous 1.8.x releases.
    Votes:
      +1: stefan2
      -1: ivan (Significant performance degradation for svnadmin hotcopy,
                the root cause is not fixed. Discussion:
                http://svn.haxx.se/dev/archive-2014-01/0158.shtml)

 Approved changes:
 =================
	* * * * * * * * * * * * * * * * * * * * * * * * * * * *
	* *
	* THIS RELEASE STREAM IS OPEN TO BUG FIXES. *
	* *
	* * * * * * * * * * * * * * * * * * * * * * * * * * * *

	This file tracks the status of releases in the 1.8.x line.

	See http://subversion.apache.org/docs/community-guide/releasing.html#release-stabilization
	for details on how release lines and voting work, what kinds of bugs can
	delay a release, etc.

	Status of 1.8.9:

	Candidate changes:
	==================

	* r1469519, r1469520, r1481848, r1481870
	Silence compiler warnings, tighten scopes and improve code.
	Justification:
	Clean builds make it more likely to notice real compiler warnings.
	Branch: ^/subversion/branches/1.8.x-r1469519-and-friends
	Votes:
	+1: blair, brane
	-0: philip (see dev thread: false alarm, trades one warning for another)

	* r1501293, r1501338, r1501360, r1501533
	Fix issue #4390, "parallel nested checkout not possible with wc-ng"
	Justification:
	Regression from 1.6. User complained. See issue for details.
	Branch:
	^/subversion/branches/1.8.x-issue4390
	Depends:
	Nomination conflicts with the branch and merges cleanly without it.
	Branch should probably be removed from the nomination (it was only needed
	before r1503552).
	Votes:
	+1: stsp
	-0: ivan, brane (not regression from 1.7.x,
	destructive change in libsvn_wc core)
	-0: rhuijben (temporarily remove my +1 until I complete investigating
	a related problem. It looks like this fix doesn't work for
	the user scenario it was intended for)

	* r1531014, r1532316, r1536350, r1536383, r1536464, r1536488, r1567204, r1567228
	Fix working copy database inconsistency on certain move scenarios,
	including those described in issue #4437 "Deleting a moved file can maintain a
	half move"
	Justification:
	The working copy database should be consistent on any normal operation.
	Branch:
	^/subversion/branches/1.8.x-issue4437
	Votes:
	+1: rhuijben
	+1: (before branch, without r1567204, r1567228) philip

	* r1518942, r1536914, r1536931
	Fix commit of a specific mixed revision copy with non copy descendants that
	shadow not present nodes. This problem manifests itself as a copy of a tree
	that can't be committed because it is reported out of date.
	Justification:
	'svn cp TREE NEW-TREE; svn ci NEW-TREE' should never report out of date
	on some random descendant.
	Branch: ^/subversion/branches/1.8.x-r1536931
	Votes:
	+1: rhuijben

	* r1536854
	Make 'svnadmin verify' detect inconsistencies that will prevent loading
	dump files.
	Justification:
	Some users rely on dump files as a means of repository backup. Without
	this patch, there is no way except of 'svnadmin load' to know that these
	dump files will load at all. With this patch, a successful verify run
	should guarantee loadable dump files.
	Branch: ^/subversion/branches/1.8.x-r1536854
	Votes:
	+1: stefan2

	* r1538812, r1538925, r1538928, r1538939
	Properly remove moved_to information when the node in its original location
	is removed.
	Justification:
	The NODES table in the working copy database should always be consistent.
	Notes:
	rhuijben: The test for this problem might accidentally depend on other
	move fixes on trunk.
	Votes:
	+1: rhuijben, philip

	* r1542610, r1564576, r1568180
	Add missing description to build.conf, to properly generate a version
	resource (including icon) for this executable on Windows.
	Notes:
	r1564576 breaks compilation on Visual Studio 2010, 2012 and 2013 of
	this project. r1568180 fixes this problem.
	Votes:
	+1: rhuijben, breser
	+1: (without r1564567, r1568180): stsp
	-0: brane (horrible grammar; suggest translation to English)

	* r1544597, r1544600, r1544688, r1544690, r1545111, r1545122
	Fix a specific case where the api behind status would report a different
	move status for A depending on whether the user called 'svn status A' or
	'svn status wc' and then looked at 'wc/A'.
	Justification:
	Api users such as AnkhSVN rely on the result of svn status to match and
	this breaks their assumptions.
	This change has the nice side effect that it resolves the most common
	user triggerable assertion in AnkhSVN, which is really caused by a
	broken move. (See other nomination for a fix). But even then it shouldn't
	matter how status is called: on the node or an ancestor.
	(It also improves status performance a tiny bit, by reducing the number
	of sqlite transactions used)
	Notes:
	r1564587 fixes a function reference on the branch in some code that
	is disabled by the macro processor on Windows.
	Branch:
	^/subversion/branches/1.8.x-r1544597
	Votes:
	+1: rhuijben
	-0 (before branch): philip (Doesn't build, needs a branch due to r1539279)

	* r1555491
	Fix svn_fs_close_root for FSFS transaction roots.
	Justification:
	Without this patch, cache-related pool cleanups may be run after the
	svn_fs_t object has been destroyed. I.e. even with explicit calls to
	svn_fs_close_root, the user may see a segfault.
	Branch:
	^/subversion/branches/1.8.x-r1555491
	Votes:
	+1: stefan2
	-0: philip, rhuijben (Seems to be a different name for the same pool?)

	* r1547454
	In FSFS, omit config file when creating pre-1.5 format repositories.
	Justification:
	Old SVN versions won't preserve this file in hotcopy. Later versions
	might also behave inconsistently in the presence of this file if their
	feature availability check is incomplete.
	Votes:
	+1: stefan2
	-0: rhuijben (Might require r1561703 to avoid breaking tests)

	* r1561426
	In FSFS, skip the generation of uniquifiers if rep-shared is no supported.
	Branch: ^/subversion/branches/1.8.x-r1561426
	Justification:
	This is not known to be a bug, atm, as we omit these when we write the
	final revisions. But we rely on those uniquifiers to be NULL in old
	repo formats when we compare reps. Them being non-NULL in a txn might
	cause problems with future changes.
	Votes:
	+1: stefan2

	* r1567996, r1568009, r1568062, r1568063
	Fix a caching inefficiency in multi-threaded servers.
	Justification:
	.
	Notes:
	Due to going back and forth with different approaches to solve this
	issue, the best way to review this
	Branch:
	^/subversion/branches/1.8.x-r1567996
	Votes:
	+1: stefan2
	-0: rhuijben (incomplete (misses cache->total_reads accesses) and
	overkill. The whole thing can be implemented with simple
	atomic operations while mutexes may be a limited resource)

	* r1571747, r1571787, r1571795
	Resolve the 'commit_util.c' (repos_root_url && repos_relpath) assertion.
	Notes:
	This assertion is raised when specifically committing a deleted
	descendant, like via a checkbox list in a GUI client.
	Justification:
	Often reported assertion by users. Upto now very hard to reproduce
	for these users. Probably because they didn's see that the node they
	deleted wasn't added to the repository yet.
	Votes:
	+1: rhuijben

	* r1572102, r1572200
	Fix a failed wc upgrade from 1.7 with exclusive wc locking.
	Notes:
	1.8 doesn't support running the regression test with exclusive locking
	so the regression test always passes.
	Justification:
	Reported by a user, upgrade is expected to work.
	Votes:
	+1: philip, brane

	* r1573088
	Add Mavericks to our sysinfo output.
	Justification:
	It's obvious, keeps us consistent with previous version behavior.
	Votes:
	+1: breser, brane

	* r1575270, r1575284
	Fix inconsistent output in svn_client_proplist4 when using absolute paths.
	Notes:
	r1575284 is the test that exposes this bug.
	Justification:
	It's a regression in 1.8.8 compared to 1.8.5.
	Votes:
	+1: brane, rhuijben

	* r1574868
	Don't let invalid mergeinfo stop 'svnadmin dump' from producing a dump.
	Part of issue #4476 "Mergeinfo containing r0 makes svnsync and svnadmin
	dump fail".
	Justification:
	A formatting error in what is essentially client-side metadata, albeit
	with repository-side interpretation as well, should never prevent
	dumping the repository. This issue only occurs when svnadmin is trying
	to warn us of mergeinfo revisions that might be unexpected, so it is
	not essential.
	Notes:
	There is a work-around, at least in theory: a dump starting from
	revision 0 or 1 does not suffer from this problem.
	Branch:
	^/subversion/branches/1.8.x-r1574868
	Votes:
	+1: julianfoad

	* r1575525, r1575915
	Resolve issue #4427: File externals can't be exported.
	Justification:
	File externals 'are just files', so why should they be discriminated
	by export?
	Notes:
	r1575525 resolves the most important issues, but in some cases still
	doesn't allow exporting file externals. r1575915 simplifies the fix and
	resolves a few remaining issues.
	Votes:
	+1: rhuijben

	* r1577151
	Make 'svn merge' honor the 'preserved-conflict-file-exts' setting.
	Justification:
	While really a behavior change that doesn't belong in a patch release
	I'm guessing nobody knew that we didn't implement this flag in this
	scenario.
	Votes:
	+1: rhuijben

	* r1577294
	Make interactive conflict resolver behave as designed.
	Justification:
	This was intended to be the behavior form the start. A typo in a
	refactoring broke it. This fix helps users avoid accidentally choosing
	'mark resolved' and committing conflict markers.
	Votes:
	+1: breser

	* r1577812, r1577813
	Resolve a segfault when 'svn merge --force' merges a directory delete
	Justification:
	Even though this merge flag is not recommended, it shouldn't cause a
	segfault.
	Notes:
	r1577812 contains the actual fix and a regression test. r1577813 is
	a minor improvement. A branch is necessary because a comment right above
	the fix contained a typo that was fixed on trunk in a huge commit of many
	typo fixes.
	Branch:
	^/subversion/branches/1.8.x-r1577812
	Votes:
	+1: rhuijben

	* r1578273, r1578311, r1578326
	Make svn_ra_get_locks() and svn_ra_get_lock() report not locked nodes
	with a NULL svn_lock_t *, as documented.
	Justification:
	Many clients use the existance of an svn_lock_t on a node via info/status
	as a boolean to note that there is a lock. But because we didn't properly
	check the result we reported mostly empty svn_lock_t instances in more
	cases. Even on directories!
	Notes:
	A branch is needed, because the serf lock handling code was refactored and
	split to multiple files on trunk.
	Branch:
	^/subversion/branches/1.8.x-r1578311
	Votes:
	+1: rhuijben

	* r1578670, r1578820
	Fix the order of node record headers written by svndumpfilter.
	Justification:
	Makes output match the rules documented in notes/dump-load-format.txt.
	Votes:
	+1: philip, rhuijben

	Veto-blocked changes:
	=====================

	* r1524145
	Make the 'm' conflict prompt option run external merge tool if configured.
	Justification:
	Usability fix. The current behaviour ('m' -> internal merge tool,
	'l' -> external merge tool) confuses users who try 'm' to run an
	external merge tool. See issue #4426.
	Votes:
	+1: stsp
	-1: breser (see my email to dev@, <5282FE15.40505@reser.org>)

	* r1560723, r1561419, r1561427, r1561428
	Fix the handling of pre-1.4 FSFS repositories.
	Branch: ^/subversion/branches/1.8.x-r1560723-and-friends
	Justification:
	Without this, old SVN versions won't be able to e.g. hotcopy those
	repos once 1.8.x made a commit in them. Also, fix our hotcopy code
	and make it cope with repositories broken by previous 1.8.x releases.
	Votes:
	+1: stefan2
	-1: ivan (Significant performance degradation for svnadmin hotcopy,
	the root cause is not fixed. Discussion:
	http://svn.haxx.se/dev/archive-2014-01/0158.shtml)

	Approved changes:
	=================