subversion

Checkout Tools
  • last updated 1 hour ago
Constraints
Constraints: committers
 
Constraints: files
Constraints: dates

Changeset 1184787 is being indexed.

Tweak the navigation bar, again.

* site-nav.html: Tweak a link text to correctly name the kind of

stuff we put in http://subversion.tigris.org/issues/.

* STATUS: Votey, votey.

Tweak the navigation bar.

* site-nav.html: Tweak a link text to be at the correct level of abstraction.

* publish/docs/community-guide/repro-template.sh:

Force all output to English.

* ./publish/contributing.html

(issues): Link to the windows reproduction script template.

Merge r1183263 from trunk:

* r1183263

Avoid dereferencing an error pointer that could be NULL.

Justification:

SEGV seen by some TSVN users

Votes:

+1: philip, stsp, steveking

Slightly changing the semantics of svn_mutex__init(). If the

enable_mutex parameter has been set, the function will try

to enable the the mutex only if SVN is potentially multi-threaded,

i.e. if APR_HAS_THREADS is set.

Thus, the parameter can simply be set to TRUE, if we would

need synchronization in a multi-threaded environment.

* subversion/libsvn_subr/svn_mutex.c

(svn_mutex__init): no-op if APR_HAS_THREADS isn't set

* subversion/include/private/svn_mutex.h

(svn_mutex__init): update docstring

* subversion/libsvn_subr/dso.c

(svn_dso_initialize2): adapt caller

* subversion/libsvn_fs_base/bdb/env.c

(bdb_init_cb): ditto

* subversion/libsvn_fs/fs-loader.c

(svn_fs_initialize): ditto

* subversion/tests/libsvn_subr/cache-test.c

(test_inprocess_cache_basic): adapt indirect caller

(test_membuffer_cache_basic): ditto

* STATUS: Vote for and approve r1183263.
Merge r1151854 from trunk, under the obvious fix rule.
Add revisions to the r1148652 nomination.
Nominate r1148652.
Follow-up to r1183358:

* subversion/libsvn_wc/wc_db.c

(op_delete_txn): When deleting the op-root of a copied-half of a move,

clear moved-to information at the delete-half. This way, deleting the

copied half has the same effect as reverting it.

* subversion/include/svn_sorts.h

(svn_sort__bsearch_lower_bound, svn_sort__array_insert,

svn_sort__array_delete): Add doxygen markup to docstrings. Explicitly

mark these functions as private. I don't really like the fact that

they are declared in a public header, but there is precedent for doing

so in this file, and moving them to a different header would cause

more code churn than it's worth.

* subversion/libsvn_client/copy.c

(extend_wc_mergeinfo, wc_to_repos_copy): Use svn_mergeinfo_merge2().

Auto-merge incoming deletions into locally moved directories on update.

Nodes which weren't modified post-move, and which are deleted at their

pre-move location during an update, are now automatically scheduled for

deletion at their post-move location, instead of becoming tree-conflict

victims. However, incoming deletions still cause a tree conflict at the

pre-move location if they want to delete nodes which were modified post-move.

Moved-here directories within which the only post-move modifications are

deleted children are not considered modified and will be deleted.

(This exposes a bug where deleting a moved-here node doesn't transform

the move into a normal delete at the pre-move location. This problem

will be fixed separately.)

* subversion/include/svn_wc.h

(svn_wc_conflict_reason_t): Add svn_wc_conflict_reason_moved_away_and_edited.

This new reason is needed so we can later differentiate the TC case handled

by this commit from pure move vs. move conflicts (i.e. where incoming and

local moves do not agree).

* subversion/tests/cmdline/tree_conflict_tests.py

(d_dels): Seperate out moves into ...

(d_moves): .. this, so tests can adjust their expectations for moves

and deletions differently.

(up_sw_dir_mod_onto_del, up_sw_dir_del_onto_del): As a result of the above

change, these two tests no longer expect a tree conflict for incoming

deletions into locally moved directories.

(up_sw_dir_del_onto_mod, up_sw_dir_del_onto_del,

(merge_dir_mod_onto_not_dir): Make sure test expectations do not change

yet in these test cases by adding d_moves to d_dels.

* subversion/tests/cmdline/update_tests.py

(update_moved_dir_leaf_del, update_moved_dir_edited_leaf_del): New tests.

These verify the new behaviour for files which were modified post-move.

Tests involving directories modifed post-move still need to be added,

but some manual testing has confirmed that they work, too.

* subversion/svn/tree-conflicts.c

(map_conflict_reason_h, map_conflict_reason_x): Add entries for

svn_wc_conflict_reason_moved_away_and_edited.

* subversion/libsvn_wc/update_editor.c

(dir_baton, make_dir_baton): Add MOVED_TO_PATH, for tree-conflict detection.

(open_root, open_directory): Check if the directory was moved away, and

if so store the moved-to path in the dir baton. Also let

check_tree_conflict() know that the node was moved-away, so it can

make an informed decision about whether a conflict should be flagged.

(modcheck_baton): Add IS_COPY.

(modcheck_callback): If IS_COPY is TRUE, only take post-copy modifications

into account, but not the fact that the node was copied/moved here.

(node_has_local_mods): Add IS_COPY parameter, same purpose as above.

(create_tree_conflict, open_directory, open_file): Make assertions

recognize svn_wc_conflict_reason_moved_away_and_edited.

(check_tree_conflict): Do not flag a tree conflict if an incoming delete

maps onto a moved node which was not modified post-move.

Update an assertion to include svn_wc_conflict_reason_moved_away_and_edited.

(delete_entry): Handle moved-away nodes such that check_tree_conflict()

doesn't flag them as a tree-conflict victim if applicable. If an

incoming delete of a moved-away node does not cause a tree-conflict,

schedule the node for deletion at its post-move location, and use

the node's post-move path in notifications.

* subversion/libsvn_wc/tree_conflicts.c

(svn_wc__conflict_reason_map): Add map entry for

svn_wc_conflict_reason_moved_away_and_edited.

Merge r1182115 from trunk:

* r1182115

Update references to subversion.tigris.org in INSTALL.

Justification:

"If you tell the truth you don't have to remember anything."

--- Mark Twain

Votes:

+1: danielsh, philip

* STATUS: Add r1183347 to the r1183054 group.

JavaHL: More backward compat improvements in the Prompter callbacks.

Patch by: Mat Booth <mat.booth@wandisco.com>

* subversion/bindings/javahl/src/org/tigris/subversion/javahl/SVNClient.java

(PromptUser1Wrapper): Call the correct askQuestion() and prompt() methods.

* STATUS: Vote for r1183263.

* STATUS: Vote.
* publish/packages.html

(windows): Link to a redesigned VisualSVN's landing page for Subversion

binaries. The page now clearly states which packages are delivered and

maintained by VisualSVN.

* STATUS: Nominate r1149135, adding Julian's +1 from IRC.

* STATUS: Propose r1183263
* subversion/libsvn_wc/wc_db.c

(read_info): Avoid dererencing a potentially NULL error pointer, ensure

all errors include the local relpath not just the checksum error.

* subversion/libsvn_wc/wc.h

(svn_wc__internal_file_modified_p): Remove documentation of nonexistent

parameter COMPARE_TEXTBASES.

* STATUS: Nominate r1183054.

Vote for the r1182904 group.
JavaHL: Fix a potential infinite recurssion bug in the JavaHL compat layer.

* subversion/bindings/javahl/src/org/tigris/subversion/javahl/SVNClient.java

(PromptUser1Wrapper): Don't infinite recurse on the askTrustSSLServer method,

by storing and using the appropriate version of the interface.

* publish/packages.html

(osx, windows): WANdisco does not require registration for these platforms.

* publish/packages.html

(osx, windows): WANdisco now provided vanilla client and server packages.