------------------------------------------------------------------------
r1493424 | ivan | 2013-06-16 01:54:05 +0400 (Âñ, 16 èþí 2013) | 11 lines

Follow-up to 1489203: Remove unnecessary check for branches ancestry from 
command line client. All required checks are performed in libsvn_client.

* subversion/svn/cl.h
* subversion/svn/util.c
  (svn_cl__check_related_source_and_target): Remove
* subversion/svn/merge-cmd.c
* subversion/svn/mergeinfo-cmd.c
  (run_merge, svn_cl__mergeinfo): Remove call to
   svn_cl__check_related_source_and_target().

------------------------------------------------------------------------
r1493475 | ivan | 2013-06-16 12:23:38 +0400 (Âñ, 16 èþí 2013) | 18 lines

Improve 'svn merge' performance over WAN: Use already fetched 
history-as-mergeinfo data to find youngest common ancestor in automatic 
merge.

* subversion/libsvn_client/client.h
  (svn_client__get_youngest_common_ancestor): Mention new
   svn_client__calc_youngest_common_ancestor in docstring.
  (svn_client__calc_youngest_common_ancestor): New declaration.

* subversion/libsvn_client/ra.c
  (svn_client__calc_youngest_common_ancestor): New. Extracted from 
   svn_client__get_youngest_common_ancestor().
  (svn_client__get_youngest_common_ancestor): Use
   svn_client__calc_youngest_common_ancestor().

* subversion/libsvn_client/merge.c:
  (find_automatic_merge): Use svn_client__calc_youngest_common_ancestor().

------------------------------------------------------------------------
r1493703 | rhuijben | 2013-06-17 14:04:38 +0400 (Ïí, 17 èþí 2013) | 15 lines

Perform an upgrade notification on every successfull working copy upgrade,
instead of only from pre-1.7 working copies.

This helps GUI clients to determine if the format bump was successfull.

* subversion/libsvn_wc/upgrade.c
  (svn_wc_upgrade): Update caller. Notify on format bumps withing WC-NG range.

* subversion/libsvn_wc/wc_db.c
  (svn_wc__db_bump_format): Re-order arguments to match common form. Provide
    optional output argument to tell about an actual format bump.

* subversion/libsvn_wc/wc_db.h
  (svn_wc__db_bump_format): Update prototype.

------------------------------------------------------------------------
r1493720 | danielsh | 2013-06-17 16:05:35 +0400 (Ïí, 17 èþí 2013) | 2 lines

* subversion/libsvn_ra/util.c
 (svn_ra__assert_capable_server): Fix typo in code.
------------------------------------------------------------------------
r1493951 | jcorvel | 2013-06-18 01:40:31 +0400 (Âò, 18 èþí 2013) | 4 lines

* tools/client-side/svn-bench/help-cmd.c: Use 'svn-bench' as program name,
   so 'svn-bench' prints "Type 'svn-bench help' for usage." instead of
   "Type 'svn help' for usage."

------------------------------------------------------------------------
r1494089 | stsp | 2013-06-18 13:49:14 +0400 (Âò, 18 èþí 2013) | 5 lines

Follow-up to r1491816:

* subversion/svn/file-merge.c
  (merge_chunks): Don't modify input arguments when concatenating chunks.

------------------------------------------------------------------------
r1494171 | stsp | 2013-06-18 19:22:42 +0400 (Âò, 18 èþí 2013) | 6 lines

* subversion/libsvn_wc/wc_db.c
  (svn_wc__db_bump_format): Keep *bumped_format as FALSE in case of error,
    rather than setting it based on the potentially undefined value of
    *result_format. Most likely not a functional change, since callers
    shouldn't be using output when this function returns an error.

------------------------------------------------------------------------
r1494223 | danielsh | 2013-06-18 21:37:28 +0400 (Âò, 18 èþí 2013) | 6 lines

Disallow 'svnadmin create --fs-type=fsfs --compatible-version=1.0', since 1.0
did not have FSFS.

* subversion/svnadmin/svnadmin.c
  (subcommand_create): Add validation.
    Note that part of the validation hard-codes the library's default fs-type.
------------------------------------------------------------------------
r1494287 | stefan2 | 2013-06-19 00:32:16 +0400 (Ñð, 19 èþí 2013) | 5 lines

Fix a "division by 0" error when upgrading a non-shared repository from
a format that supports sharding.

* subversion/libsvn_fs_fs/fs_fs.c
  (upgrade_body): pack revprop shards for sharded repositories only
------------------------------------------------------------------------
r1494298 | stefan2 | 2013-06-19 00:59:07 +0400 (Ñð, 19 èþí 2013) | 12 lines

Fix a data loss when 'svnadmin upgrade' gets interrupted after packing the
revprops but before bumping the format file.  In that case, the non-packed
revprops would have already been deleted but the next 'svnadmin upgrade'
would drop the packed shard as well and then fail to repack.

The fix is to defer the cleanup until after the format bump.

* subversion/libsvn_fs_fs/fs_fs.c
  (upgrade_pack_revprops): don't remove the non-packed shards here ...
  (upgrade_cleanup_pack_revprops): ... but in this new function
  (upgrade_body): call the cleanup after the format bump

------------------------------------------------------------------------
r1494318 | stefan2 | 2013-06-19 01:32:04 +0400 (Ñð, 19 èþí 2013) | 4 lines

Follow-up to r1494298: actually call the cleanup function

* subversion/libsvn_fs_fs/fs_fs.c
  (upgrade_body): fix copy'n'pasto
------------------------------------------------------------------------
r1494342 | danielsh | 2013-06-19 02:17:09 +0400 (Ñð, 19 èþí 2013) | 3 lines

* subversion/tests/svn_test_fs.c
  (svn_hash.h): Include.
  (make_fs_config): Add SVN_FS_CONFIG_PRE_1_8_COMPATIBLE to the config hash.
------------------------------------------------------------------------
r1494657 | stsp | 2013-06-19 19:33:10 +0400 (Ñð, 19 èþí 2013) | 8 lines

Remove duplicate invocations of the conflict resolver from the update editor.

* subversion/libsvn_wc/update_editor.c
  (mark_directory_edited, mark_file_edited): Don't invoke the conflict
   resolver here, the close_file() and close_directory() functions take
   care of this already.


------------------------------------------------------------------------
r1494913 | stefan2 | 2013-06-20 14:31:53 +0400 (×ò, 20 èþí 2013) | 7 lines

Fix a severe performance regression in 'svn log':  The new log code
will run svn_client__repos_location_segments even if we log on the
root folder.  That triggered a long-standing performance issue in
fs_node_origin_rev.  This patch fixes the latter.

* subversion/libsvn_fs_fs/tree.c
  (fs_node_origin_rev): special case root node IDs instead of walking
------------------------------------------------------------------------
r1494967 | stefan2 | 2013-06-20 17:02:07 +0400 (×ò, 20 èþí 2013) | 4 lines

Attempt a similar fix to BDB as r1494913 provided for FSFS.

* subversion/libsvn_fs_base/tree.c
  (base_node_origin_rev): add efficient shortcut for the root folder
------------------------------------------------------------------------
r1495063 | stsp | 2013-06-20 19:53:32 +0400 (×ò, 20 èþí 2013) | 9 lines

* subversion/libsvn_fs_fs/tree.c
  (cache_lookup): Calculate hash_value in an alternative way on
  strict-alignment architectures, such as Solaris/sparc. Wrap the current
  calculation in SVN_UNALIGNED_ACCESS_IS_OK to ensure that it is only
  used on platforms which support unaligned access.

Found by: Rainer Jung <rjung@apache.org>
Thanks to Sergey Raevskiy for spotting a bug in my initial patch.

------------------------------------------------------------------------
r1495104 | rhuijben | 2013-06-20 21:06:22 +0400 (×ò, 20 èþí 2013) | 71 lines

Improve http status checking in ra_serf, by providing a default error for
unexpected server error, unexpected client error and unexpected moved status
values instead of handling all of the unexpected values as SVN_NO_ERROR.

This turns the
[[
In file
   'D:\Development\SVN\Releases\TortoiseSVN-1.8.0\ext\subversion\subversion\libsvn_client\ra.c'
   line 647: assertion failed (peg_revnum != SVN_INVALID_REVNUM)
]]

into the http error that is causing this problem.
Something like:
[[
svn: E175002: DAV request failed: Content length required
]]

* subversion/libsvn_ra_serf/blame.c
  (svn_ra_serf__get_file_revs): Update caller.

* subversion/libsvn_ra_serf/commit.c
  (return_response_err): Update caller.

* subversion/libsvn_ra_serf/getdate.c
  (svn_ra_serf__get_dated_revision): Add status check and verify result to
    make sense.

* subversion/libsvn_ra_serf/getlocations.c
  (svn_ra_serf__get_locations): Update caller.

* subversion/libsvn_ra_serf/getlocationsegments.c
  (svn_ra_serf__get_location_segments): Update caller.

* subversion/libsvn_ra_serf/getlocks.c
  (svn_ra_serf__get_locks): Update caller.

* subversion/libsvn_ra_serf/inherited_props.c
  (svn_ra_serf__get_inherited_props): Update caller.

* subversion/libsvn_ra_serf/log.c
  (svn_ra_serf__get_log): Update caller.

* subversion/libsvn_ra_serf/mergeinfo.c
  (svn_ra_serf__get_mergeinfo): Update caller. Don't drop detailed errors.

* subversion/libsvn_ra_serf/options.c
  (svn_ra_serf__v2_get_youngest_revnum): Update caller. Verify if result
    makes sense.
  (svn_ra_serf__v1_get_activity_collection): Update caller.
  (svn_ra_serf__exchange_capabilities): Update caller.

* subversion/libsvn_ra_serf/property.c
  (svn_ra_serf__wait_for_props): Update caller.

* subversion/libsvn_ra_serf/ra_serf.h
  (svn_ra_serf__error_on_status): Update argument to pass the entire status to
    allow creating better messages for unknown errors.

* subversion/libsvn_ra_serf/replay.c
  (svn_ra_serf__replay): Update caller.
  (svn_ra_serf__replay_range): Update caller.

* subversion/libsvn_ra_serf/update.c
  (handle_stream): Update caller.
  (finish_report): Update caller.

* subversion/libsvn_ra_serf/util.c
  (svn_ra_serf__error_on_status): Extend arguments. Provide specialized error
    for http status 411 and a generic error for all currently unexpected
    errors.

------------------------------------------------------------------------
r1495204 | stefan2 | 2013-06-21 01:40:50 +0400 (Ïò, 21 èþí 2013) | 16 lines

Follow-up to r1495063: integer overflows result in an inefficient hash
function and reduce cache effectiveness.

This illustrates what happened before the patch:

  char c = 99;
  unsigned hash = 0;
  hash |= c << 8; /* c << 8 is often 0, actually it's undefined */
 
On a more general note, you don't want to make it easy for parts of
the input to cancel each other out.  So, adding (potentially) negative
values is a bad thing (strategically).

* subversion/libsvn_fs_fs/tree.c
  (cache_lookup): prevent unnecessary integer overflows in hash function

------------------------------------------------------------------------
r1495209 | stefan2 | 2013-06-21 02:11:31 +0400 (Ïò, 21 èþí 2013) | 10 lines

Make the hash function used by the FSFS DAG node cache
platform-independent. That should help us reproducing issues
detected on "exotic" platforms.

* subversion/libsvn_fs_fs/tree.c
  (cache_lookup): normalize chunked calculation to big endian

Patch by: stsp

Reverted in r1495256.
------------------------------------------------------------------------
r1495214 | stsp | 2013-06-21 02:17:26 +0400 (Ïò, 21 èþí 2013) | 5 lines

Follow-up to r1495209:

* subversion/libsvn_fs_fs/tree.c: Include apr_want.h for htonl().

Reverted in r1495256.
------------------------------------------------------------------------
r1495256 | stefan2 | 2013-06-21 04:36:48 +0400 (Ïò, 21 èþí 2013) | 5 lines

Effectively revert r1495209 and r1495214.  Tune and document the remainder.

* subversion/libsvn_fs_fs/tree.c
  (cache_lookup): turn magic number into a constant;
                  fine-tune iteration latencies; add commentary
------------------------------------------------------------------------
r1495329 | philip | 2013-06-21 12:35:37 +0400 (Ïò, 21 èþí 2013) | 9 lines

Fix the svnmucc symlink in svn-tools so it doesn't to link to DESTDIR.
The user can configure $(bindir) but $(toolsdir) is always a subdir
of that so a relative symlink should be OK.

Found by: Nico Kadel-Garcia <nkadel{_AT_}gmail.com>

* Makefile.in
  (INSTALL_EXTRA_TOOLS): Use a relative symlink.

------------------------------------------------------------------------
r1495428 | danielsh | 2013-06-21 17:25:32 +0400 (Ïò, 21 èþí 2013) | 5 lines

* subversion/libsvn_fs_base/bdb/env.c
  (bdb_error_gatherer): Improve the default error code.

    This was motivated by the following error (on users@):
    svnadmin: E000000: bdb: Program version 5.1 doesn't match environment version 4.4
------------------------------------------------------------------------
r1495432 | danielsh | 2013-06-21 17:37:37 +0400 (Ïò, 21 èþí 2013) | 7 lines

* subversion/libsvn_repos/fs-wrap.c
  (svn_repos__validate_prop): Consider NULL a valid value, to allow
    deleting non-regular props from historical repositories that have them.

Suggested by: Arwin Arni
(along with a patch)

------------------------------------------------------------------------
r1495446 | danielsh | 2013-06-21 18:17:41 +0400 (Ïò, 21 èþí 2013) | 2 lines

* subversion/libsvn_repos/fs-wrap.c
  (svn_repos__validate_prop): Clarify a comment.
------------------------------------------------------------------------
r1495805 | stefan2 | 2013-06-23 14:36:19 +0400 (Âñ, 23 èþí 2013) | 7 lines

Simplify the hash function used for the DAG node cache.

* subversion/libsvn_fs_fs/tree.c
  (cache_entry_t): peg hash_value to unsigned 32 bits which minimizes
                   the number of casts we have to do
  (cache_lookup): same here, including FACTOR;
                  drop the chunky alternative to unaligned data access
------------------------------------------------------------------------
r1495850 | lgo | 2013-06-23 21:14:58 +0400 (Âñ, 23 èþí 2013) | 8 lines

Fix a crash in TSVN when receiving a property update on a directory with local
property modifications.
(See: http://svn.haxx.se/dev/archive-2013-06/0543.shtml)

* subversion/libsvn_wc/update_editor.c
  (close_directory): Use eb->cancel_baton instead of eb->conflict_baton for the
      cancel_baton parameter in the call to svn_wc__conflict_invoke_resolver.

------------------------------------------------------------------------
r1495978 | danielsh | 2013-06-24 14:15:31 +0400 (Ïí, 24 èþí 2013) | 4 lines

Silence a compiler warning.

* subversion/libsvn_fs_fs/tree.c
  (cache_lookup): Only declare FACTOR when it is used.
------------------------------------------------------------------------
r1496110 | pburba | 2013-06-24 20:15:27 +0400 (Ïí, 24 èþí 2013) | 6 lines

Avoid a segfault in svn_client_log5, see
http://svn.haxx.se/dev/archive-2013-06/0522.shtml

* subversion/libsvn_client/log.c
  (run_ra_get_log): bsearch can return NULL, so handle that.

------------------------------------------------------------------------
r1496132 | rhuijben | 2013-06-24 21:40:21 +0400 (Ïí, 24 èþí 2013) | 11 lines

Provide a slightly more detectable error code than APR_EGENERAL when a
connection fails on opening an ra session.

The combination serf/apr should really provide the error stored
for the connection (in the sockopt SO_ERROR) instead of this catch all error,
to allow users and clients access to more detailed diagnosis.

* subversion/libsvn_ra_serf/serf.c
  (svn_ra_serf__open): Wrap APR_EGENERAL with an error that is diagnosable
    from code as being network related.

------------------------------------------------------------------------
r1496151 | rhuijben | 2013-06-24 22:29:34 +0400 (Ïí, 24 èþí 2013) | 3 lines

* subversion/libsvn_ra_serf/serf.c
  (svn_ra_serf__open): Following up on r1496132, don't crash on non errors.

------------------------------------------------------------------------
r1496938 | julianfoad | 2013-06-26 17:57:47 +0400 (Ñð, 26 èþí 2013) | 2 lines

* subversion/svn/svn.c
  (svn_cl__cmd_table): Tweak grammar in merge help text, following r1491432.
------------------------------------------------------------------------
r1496957 | stefan2 | 2013-06-26 19:14:43 +0400 (Ñð, 26 èþí 2013) | 4 lines

Try a different approach to optimize 'svn log' on the repo root.

* subversion/libsvn_client/log.c
  (svn_client_log5): at repo root, we already know the only location segment
------------------------------------------------------------------------
r1497002 | philip | 2013-06-26 20:50:59 +0400 (Ñð, 26 èþí 2013) | 17 lines

When possible pass the stream's file to external diff rather than
creating a temporary spool file.  Fixes issue 4382.

* subversion/include/private/svn_io_private.h
  (svn_stream__aprfile): New.

* subversion/libsvn_subr/stream.c
  (struct svn_stream_t): Add file member.
  (svn_stream_create, svn_stream_from_aprfile2): Set file member.
  (svn_stream__aprfile): New.

* subversion/libsvn_client/diff.c
  (diff_content_changed): Use stream's file if possible.

* subversion/svnlook/svnlook.c
  (print_diff_tree): Use stream's file if possible.

------------------------------------------------------------------------
r1497318 | stsp | 2013-06-27 16:01:33 +0400 (×ò, 27 èþí 2013) | 8 lines

Better explain working copy locks in 'svn help status' output.

* subversion/svn/svn.c
  (svn_cl__cmd_table): When describing the third status cloumn, which shows
   'L', clearly state the purpose of the working copy write lock. This should
   avoid confusing with other kinds of locks, such as the ones created with
   the 'svn lock' command.

------------------------------------------------------------------------
r1497319 | stsp | 2013-06-27 16:11:21 +0400 (×ò, 27 èþí 2013) | 8 lines

Better explain exclusive-commit locks in 'svn help status' output.

* subversion/svn/svn.c
  (svn_cl__cmd_table): When describing the sixth status cloumn, which shows
   local and remove lock tocken state, use wording that is easier to understand
   for novice users, who might not exactly know what a 'lock token' is, for
   example.

------------------------------------------------------------------------
r1497551 | gstein | 2013-06-28 00:09:51 +0400 (Ïò, 28 èþí 2013) | 7 lines

Fix the Accept-Encoding header format.

Reported by: George Schizas <gschizas{_AT_}gmail.com>

* subversion/libsvn_ra_serf/update.c:
  (setup_update_report_headers): use "," between format types

------------------------------------------------------------------------
r1497804 | danielsh | 2013-06-28 19:13:35 +0400 (Ïò, 28 èþí 2013) | 8 lines

Tweak sqlite error messages: put the error code consistently next to the
"sqlite:" prefix (except one place where there is no such prefix).

* subversion/libsvn_subr/sqlite.c
  (SQLITE_ERR, SQLITE_ERR_MSG, exec_sql2, svn_sqlite__step):
    Use 'sqlite[S%d]:' prefix.
  (init_sqlite):
    Use square brackets around 'S%d' for consistency.
------------------------------------------------------------------------
r1498136 | brane | 2013-06-30 18:42:05 +0400 (Âñ, 30 èþí 2013) | 11 lines

Disable named atomics if APR does not implement memory-mapped files.
People still use Subversion on platforms where mmap is not available.

See: http://mail-archives.apache.org/mod_mbox/subversion-dev/201306.mbox/%3C51CE9626.2000201%40gknw.net%3E

* subversion/libsvn_subr/named_atomic.c
  (mutex_initialized, init_thread_mutex, delete_lock_file):
   Elide when !APR_HAS_MMAP.
  (svn_named_atomic__is_supported): Return FALSE when !APR_HAS_MMAP.
  (svn_atomic_namespace__create): Return APR_ENOTIMPL when !APR_HAS_MMAP.

------------------------------------------------------------------------
r1498449 | danielsh | 2013-07-01 17:47:14 +0400 (Ïí, 01 èþë 2013) | 5 lines

Break out a helper function.

* subversion/libsvn_ra/util.c
  (svn_ra__assert_mergeinfo_capable_server): Break out some code into..
  (get_path): .. this new function.  While here, mark a string for translation.
------------------------------------------------------------------------
r1498455 | danielsh | 2013-07-01 17:51:26 +0400 (Ïí, 01 èþë 2013) | 4 lines

Fix a segfault in kidney blame.

* subversion/libsvn_ra/util.c
  (svn_ra__assert_capable_server): Handle NULL paths.
------------------------------------------------------------------------
r1498456 | danielsh | 2013-07-01 17:52:50 +0400 (Ïí, 01 èþë 2013) | 5 lines

ra_serf: when querying for the 'get-file-revs-reverse' capability, answer
'No' rather than 'Never heard of that one.'

* subversion/libsvn_ra_serf/options.c
  (options_response_handler): Recognise SVN_RA_CAPABILITY_GET_FILE_REVS_REVERSE.
------------------------------------------------------------------------
r1498483 | danielsh | 2013-07-01 18:35:37 +0400 (Ïí, 01 èþë 2013) | 7 lines

svn_test: Add a helper function.

* subversion/tests/svn_test_fs.h
  (svn_test__validate_tree): Declare.

* subversion/tests/svn_test_fs.c
  (svn_test__validate_tree): Implement.
------------------------------------------------------------------------
r1498484 | danielsh | 2013-07-01 18:36:40 +0400 (Ïí, 01 èþë 2013) | 8 lines

Improve a regression test.

* subversion/tests/libsvn_fs/fs-test.c
  (svn_hash.h): Include.
  (filename_trailing_newline): Switch from a blacklist approach to
    to a whitelist approach, for defining backends that don't implement
    the API correctly.  Also extend the test to actually commit the files and
    dirs, and verify the commit.
------------------------------------------------------------------------
r1498486 | danielsh | 2013-07-01 18:38:27 +0400 (Ïí, 01 èþë 2013) | 4 lines

* subversion/tests/libsvn_fs/fs-test.c
  (filename_trailing_newline): Rename a local variable.

Suggested by: stsp
------------------------------------------------------------------------
r1498550 | danielsh | 2013-07-01 20:27:33 +0400 (Ïí, 01 èþë 2013) | 7 lines

Fix a client segfault in svn_ra_get_commit_editor3() with a NULL log message.

* subversion/libsvn_ra_svn/client.c
  (ra_svn_commit): Handle a NULL log message.

* subversion/libsvn_ra_svn/protocol
  (commit): Note an API change to do when next revving this protocol command.
------------------------------------------------------------------------
r1498564 | rhuijben | 2013-07-01 20:53:35 +0400 (Ïí, 01 èþë 2013) | 15 lines

Allow meta-data only moves to succeed when there is already an unversioned
file as the move target.

This 'fix-after-move' scenario is exactly one of the scenarios where this
option was added for.

Found via AnkhSVN user issue.

* subversion/libsvn_client/copy.c
  (verify_wc_dsts): Add metadata_only argument and don't check for local
    unversioned obstructions if it is set.
  (verify_wc_srcs_and_dsts): Add metadata_only argument and update caller.
  (repos_to_wc_copy_locked): Update caller.
  (try_copy): Update caller.

------------------------------------------------------------------------
r1498851 | rhuijben | 2013-07-02 13:16:55 +0400 (Âò, 02 èþë 2013) | 13 lines

In the serf transition based xml parser: provide an explicit error if the root
element isn't matched in the transition table.

Without this most parsers just return success for every invalid element such
as '<html>' when the file contains valid xml.

This requires a better (new) error code which I'll add as a separate commit, to
allow backporting this.

* subversion/libsvn_ra_serf/xml.c
  (svn_ra_serf__xml_cb_start): Expect transitions from the default state to
    succeed, by returning an error when it doesn't.

------------------------------------------------------------------------
r1498885 | brane | 2013-07-02 15:37:54 +0400 (Âò, 02 èþë 2013) | 7 lines

Fix change initialization in the Ev2 compatibility shims.
The shims expect an uninitialized change record's kind to be svn_node_unknown,
but it was initialized to zero, which is svn_node_none.

* subversion/libsvn_delta/compat.c:
  (locate_change): Init change->kind explicitly to svn_node_unknown.

------------------------------------------------------------------------
r1498997 | stsp | 2013-07-02 20:18:30 +0400 (Âò, 02 èþë 2013) | 4 lines

* subversion/libsvn_ra/util.c
  (svn_ra__assert_capable_server): Only call get_path() in error case, it's
   not needed otherwise.

------------------------------------------------------------------------
r1499034 | danielsh | 2013-07-02 21:44:58 +0400 (Âò, 02 èþë 2013) | 4 lines

* build/ac-macros/sqlite.m4
  (SVN_LIB_SQLITE.--with-sqlite): Look for sqlite-amalgamation in the build
    dir, too.
  (SVN_DOWNLOAD_SQLITE): Adjust error message accordingly.
------------------------------------------------------------------------
r1499064 | stsp | 2013-07-02 23:12:34 +0400 (Âò, 02 èþë 2013) | 5 lines

Fix issue #4387: Regression: svn merge --record-only errors out on renamed path

* subversion/libsvn_client/merge.c
  (record_tree_conflict): Don't record a conflict during record-only merges.

------------------------------------------------------------------------
r1499095 | stsp | 2013-07-03 00:19:10 +0400 (Ñð, 03 èþë 2013) | 6 lines

Fix issue #4388, crash during tree conflict resolution.

* subversion/libsvn_wc/wc_db_update_move.c
  (svn_wc__db_resolve_delete_raise_moved_away): Don't deref old_version
   unconditionally, as it can be NULL.

------------------------------------------------------------------------
r1499096 | stsp | 2013-07-03 00:22:26 +0400 (Ñð, 03 èþë 2013) | 6 lines

* subversion/libsvn_wc/wc_db_update_move.c
   (svn_wc__db_update_moved_away_conflict_victim): Pre-emptively guard
    against NULL deref in case either or both old_version and new_version
    are NULL. Found by code inspection, not sure if a crash can be triggered.
    See also r1499095 and issue #4388.

------------------------------------------------------------------------
r1499100 | breser | 2013-07-03 00:31:57 +0400 (Ñð, 03 èþë 2013) | 6 lines

Add test for config files with UTF-8 Byte Order Mark.

* subversion/tests/libsvn_sub/config-test.c
  (test_ignore_bom, test_funcs): Add new test.
  

------------------------------------------------------------------------
r1499403 | danielsh | 2013-07-03 17:48:57 +0400 (Ñð, 03 èþë 2013) | 4 lines

Follow-up to r1495329:

* Makefile.in
  (INSTALL_EXTRA_TOOLS): Use an absolute symlink instead of a relative 
------------------------------------------------------------------------
r1499438 | danielsh | 2013-07-03 19:17:07 +0400 (Ñð, 03 èþë 2013) | 14 lines

Block building with BDB 6.

* build/ac-macros/berkeley-db.m4
  (SVN_LIB_BERKELEY_DB_TRY): Block building with BDB 6, with a comment
    explaining why.
  (SVN_LIB_BERKELEY_DB): Document that only 4.x and 5.x are supported.
  (--with-berkeley-db): Ditto.  The help string does not evaluate the $,
    but that is a preexisting bug that should be fixed separately.

* build/generator/gen_win.py
  (GeneratorBase._find_bdb): Add a comment cautioning against enabling
    BDB 6.

* configure.ac: Define $db_alt_version and use it in an error message.
------------------------------------------------------------------------
r1499447 | danielsh | 2013-07-03 19:43:12 +0400 (Ñð, 03 èþë 2013) | 4 lines

* configure.ac: Nuke the warning at the end about not building with BDB,
    because the BDB backend is deprecated.
    The warning had a bug: it referred to $with_berkeley_db which is never set.

------------------------------------------------------------------------
r1499460 | danielsh | 2013-07-03 20:20:21 +0400 (Ñð, 03 èþë 2013) | 12 lines

configure: Add an opt-in for BDB 6.

Review by: stsp
(He co-authored the warning text, too.)

* build/ac-macros/berkeley-db.m4
  (--enable-banging-head-against-wall): New option.
  (SVN_LIB_BERKELEY_DB_TRY): Grow new parameter, add a possible return value.
  (SVN_LIB_BERKELEY_DB): Add a new failure mode when bdb6 was found but the
    opt-in not used.

* configure.ac: Add a detailed warning message for "Found, but not opt-in" case.
------------------------------------------------------------------------
r1499483 | breser | 2013-07-03 21:17:24 +0400 (Ñð, 03 èþë 2013) | 5 lines

Fix a typo in the config file BOM test.

* subversion/tests/libsvn_subr/config-test.c
  (test_ignore_bom): First char of UTF-8 BOM is 0xEF not 0xEE

------------------------------------------------------------------------
r1499492 | breser | 2013-07-03 21:30:21 +0400 (Ñð, 03 èþë 2013) | 8 lines

Fix incorrect conversion of char to int in config file parsing.

See: https://www.securecoding.cert.org/confluence/display/seccode/STR34-C.+Cast+characters+to+unsigned+char+before+converting+to+larger+integer+sizes

* subversion/libsvn_subr/config_file.c
  (parser_getc, parser_getc_plain): Before converting char to int cast to
    unsigned char.

------------------------------------------------------------------------
r1499496 | breser | 2013-07-03 21:42:42 +0400 (Ñð, 03 èþë 2013) | 9 lines

Allow config files to have a UTF-8 BOM at the start.

* subversion/libsvn_subr/config_file.c
  (skip_bom): New function to skip a UTF-8 BOM.
  (svn_config__parse_stream): Call skip_bom() at start of parsing.

* subversion/tests/libsvn_subr/config-test.c
  (test_funcs): Mark test_ignore_bom as passing.

------------------------------------------------------------------------
r1499498 | breser | 2013-07-03 21:50:46 +0400 (Ñð, 03 èþë 2013) | 6 lines

Clean up some style nits in skip_bom().

* subversion/libsvn_subr/config_file.c
  (skip_bom): Remove some unnecessary syntax that I thought I'd
    removed before committing r1499496.

------------------------------------------------------------------------
r1499727 | danielsh | 2013-07-04 15:45:20 +0400 (×ò, 04 èþë 2013) | 18 lines

Follow-up to r1498550: widely document the implementation detail that r1498550
relies on, and elevate it to a protocol promise.

No functional change (comment changes only).

Suggested by: philip

* subversion/libsvn_ra_svn/client.c
  (ra_svn_commit): Extend documentation of the LOG_MSG manipulation.

* subversion/libsvn_ra_svn/protocol
  (commit):
    Specify the behaviour of the server when both LOGMSG and REVPROP-TABLE are
    present.  I verified (by code inspection) that 1.5.x@HEAD behaves that way,
    so presumably everything in [1.5.0, trunk@HEAD].

* subversion/svnserve/serve.c
  (commit): Point to the new protocol promise.
------------------------------------------------------------------------
r1500074 | danielsh | 2013-07-05 21:21:09 +0400 (Ïò, 05 èþë 2013) | 7 lines

Properly format the error message during unintended switches.

* subversion/libsvn_client/cmdline/switch.c
  (switch_internal): Use local style for paths when formatting the "no common
    ancestry" error message.

Patch by: Evgeny Kotkov <evgeny.kotkov{_AT_}visualsvn.com>
------------------------------------------------------------------------
r1500175 | danielsh | 2013-07-06 03:57:53 +0400 (Ñá, 06 èþë 2013) | 4 lines

Followup to r1499034:

* build/ac-macros/sqlite.m4
  (SVN_DOWNLOAD_SQLITE): Don't print the same path twice.
------------------------------------------------------------------------
r1500226 | brane | 2013-07-06 12:19:06 +0400 (Ñá, 06 èþë 2013) | 10 lines

Fix a thinko in the Ev2 shims that caused the driver to prepend a "/"
to copyfrom paths without considering that they might be URLs already.

The bug was triggered by running the tests for the Ev2 commit editor
in JavaHL via http://.

* subversion/libsvn_delta/compat.c (apply_change):
   Try to make an FS path (starting with "/") from the copyfrom path only
   when the editor baton does not contain a proper reposiutory root URL.

------------------------------------------------------------------------
r1500680 | stsp | 2013-07-08 16:07:35 +0400 (Ïí, 08 èþë 2013) | 9 lines

Use local style for paths when formatting error messages from the
'svn diff --old A --new B' commands.  Follow-up to r1310291.

* subversion/libsvn_client/diff_local.c
  (svn_client__arbitrary_nodes_diff): Use local style for paths when formatting
    the "not the same node kind" and "not a file or directory" messages.

Patch by: Evgeny Kotkov <evgeny.kotkov{_AT_}visualsvn.com>

------------------------------------------------------------------------
r1500695 | stsp | 2013-07-08 16:37:07 +0400 (Ïí, 08 èþë 2013) | 8 lines

Rename --enable-banging-head-against-wall to --enable-bdb6. I still like the
former name for this option, however using a clear name will help packagers
and people trying to read packaging scripts, and that's more important than
making a joke.

* build/ac-macros/berkeley-db.m4
  configure.ac: Rename --enable-banging-head-against-wall to --enable-bdb6.

------------------------------------------------------------------------
r1500762 | stsp | 2013-07-08 18:29:04 +0400 (Ïí, 08 èþë 2013) | 13 lines

Prevent the GPG Agent password store from overriding other password stores
even though no usable GPG Agent is running.

* subversion/libsvn_subr/gpg_agent.c
  (find_running_gpg_agent): New helper function that finds a running GPG Agent.
   This code was split out from the upper half of ...
  (password_get_gpg_agent): ... this function.
  (password_set_gpg_agent): Don't unconditionally set *done to TRUE. While it
   is correct that we don't need to explicitly store the password in the agent,
   this function must check whether a usable GPG Agent exists. Otherwise,
   users might be left with no usable password store, since the gpg-agent
   store wrongly claims that the password was saved, when it actually wasn't.

------------------------------------------------------------------------
r1500799 | stsp | 2013-07-08 19:34:55 +0400 (Ïí, 08 èþë 2013) | 7 lines

* subversion/libsvn_subr/gpg_agent.c
  (password_set_gpg_agent): Don't require GPG_TTY to be set in the environment.
   This variable is only needed in some GPG Agent configurations, and we should
   allow other configurations to work as well. For example, a GUI app might
   talk to GPG agent without a TTY. Apparently, this is the case on Mac OS X,
   according to Ben Reser.

------------------------------------------------------------------------
r1500801 | stsp | 2013-07-08 19:37:28 +0400 (Ïí, 08 èþë 2013) | 5 lines

* subversion/libsvn_subr/gpg_agent.c
  (password_get_gpg_agent): Don't require GPG_TTY and TERM to be set in the
   environment. These variables are only needed in some GPG Agent
   configurations. See also r1500799.

------------------------------------------------------------------------
r1500802 | stsp | 2013-07-08 19:41:01 +0400 (Ïí, 08 èþë 2013) | 3 lines

* subversion/libsvn_subr/gpg_agent.c
  (password_set_gpg_agent): Remove unused variable.

------------------------------------------------------------------------
r1500904 | breser | 2013-07-08 23:40:29 +0400 (Ïí, 08 èþë 2013) | 5 lines

swig-pl: Fix typo introduced in r1490721.

* subversion/bindings/swig/perl/native/Client.pm
  ($revision): s/alos/also/

------------------------------------------------------------------------
r1500928 | danielsh | 2013-07-09 00:25:22 +0400 (Âò, 09 èþë 2013) | 5 lines

* build/ac-macros/berkeley-db.m4: Change --disable-bdb6 to --enable-bdb6
    in the docstring, for consistency with every other helpstring in our
    configure.

Found by: breser