Checkout Tools
  • last updated 3 hours ago
Constraints: committers
Constraints: files
Constraints: dates

Changeset 874187 is being indexed.

* www/links.html (hosting): Fix Assembla entry.

Found by: Ryan Holbrook <>

Initialize all members when creating a svn_wc_conflict_description_t struct.

* subversion/libsvn_wc/util.c

(svn_wc_conflict_description_create_tree): Use apr_pcalloc instead

of apr_palloc to allocate memory.

* subversion/po/ko.po: More work in korean translation (133 untranslated messages remained)

* subversion/libsvn_subr/auth.c: Include svn_dso.h. Follow-up to r34085.

Deleting the issue-2843-dev branch, it was merged back to trunk in r34108.
Merge the issue-2843 branch back to trunk.

I wasn't able to use reintegrate, see So

this was done the old-fashioned way, with a 2-URL merge:

Note: This merge added mergeinfo on the subversion directory. This

mergeinfo originated in r31940 on the issue-2843 branch, but there is no

obvious reason why it was there. r31940 was a merge of r31359-31939 from

trunk, but there is no mergeinfo diff between those revs, so where this came

from is a complete mystery. In that light I removed this mergeinfo prior to

the commit.

  1. … 24 more files in changeset.
Do a --record-only merge of -r31357:34104 from trunk so this branch will

reintegrate back to trunk cleanly.

Fix issue #3313 by validating all eligible properties (and among them,

svn:author) for being valid UTF-8.

* subversion/libsvn_repos/fs-wrap.c


Apply the "is UTF-8 and LF" check whenever svn_prop_needs_translation()

is true (that is the very definition of the latter).

On the issue-2843-dev branch: bring up-to-date with latest trunk (r34104).

* subversion/tests/cmdline/

(test_list): Expect commit_out_of_date_deletions to fail over ra_svn.

I'm not sure why it wasn't marked this way originally. See r34101

and r34073 for more context; I think there was just a multi-way

confusion here and that it is now un-confused.

Found by: danielsh

(He noticed my IRC inquiries and pointed out that ra_svn failure persists.)

On the issue-2843-dev branch:

* subversion/tests/cmdline/

(test_list): Expect excluded_receive_remote_removal to pass now.

On the issue-2843-dev branch: bring up-to-date with latest trunk (r34101).

* subversion/tests/cmdline/

(test_list): Expect commit_out_of_date_deletions to pass over DAV now.

I'm not sure when it started XPASSing; it seems to have been doing

so at least since r34008, but it may have started long before then.

On the issue-2843-dev branch: resolve conflicts from the r34098 merge.

* subversion/libsvn_wc/entries.c

(read_entries): Resolve conflict between r31713 and r34093.

On the fsfs-pack branch:

Bring up-to-date with trunk.

  1. … 54 more files in changeset.
On the issue-2843-dev branch: bring up-to-date with trunk@HEAD (r34097):

svn merge

There was one conflict. I am committing it with conflict markers

intact, to be resolved in a separate revision:


  1. … 73 more files in changeset.
Add r34096 to the r34082 group, and remove note about needing extra eyes --

I got the best eyes possible on it already.

Minor follow-up to r34082.

* subversion/mod_dav_svn/mirror.c

(dav_svn__proxy_merge_fixup): Move a return statement to avoid unnecessary

processing when a GET or PROPFIND is not proxied through.

Suggested by: jerenkrantz

* subversion/mod_dav_svn/repos.c

(parse_querystring): use authz_read_func when tracing node locations.

Follow-up to r34091, adjust expected output in a merge test to work with

debug builds.

Suggested by: cmpilato

* subversion/tests/cmdline/

(reintegrate_with_subtree_mergeinfo): Debug builds also output the

file/lineno of the error, so expect that.

Fix the entries walker again: allow it to walk into a deleted-but-re-added

directory. This is a follow-up to r34084 which fixed a similar bug but broke

this case. Also factor out the test used here and in several other places to

make it easier to keep it consistent.

* subversion/libsvn_wc/entries.c

(entry_is_hidden): New function.

(handle_start_tag, read_entries): Use it. (No functional change.)

(walker_helper, visit_tc_too_found_entry): Use it, fixing the bug mentioned.

Delete the reintegrate-improvements branch, it was merged back to trunk.
Reintegrate the reintegrate-improvements branch back to trunk.
  1. … 7 more files in changeset.
On the reintegrate-improvements branch: Merge r34059-34089 from trunk.
    • ?
  1. … 36 more files in changeset.
Implement a WC entries walker like svn_wc_walk_entries3() but that also

visits unversioned tree conflict victims.

This may well want to become public, perhaps as a new revision of

svn_wc_walk_entries3(). It could do with a better name than ..._and_tc().

* subversion/libsvn_wc/wc.h

(svn_wc__walk_entries_and_tc): New function.

* subversion/libsvn_wc/entries.c

(visit_tc_too_baton_t): New type.

(visit_tc_too_callbacks): New constant.

(visit_tc_too_found_entry, visit_tc_too_error_handler,

svn_wc__walk_entries_and_tc): New functions.

Allow FSFS rep-caching to be configurable.

* subversion/libsvn_fs_fs/fs_fs.c

(write_config): Add section for rep-sharing, currently disabled by default.

(svn_fs_fs__open, svn_fs_fs__create): Check the config before opening or

creating rep cache.

* subversion/libsvn_fs_fs/fs.h


* subversion/libsvn_fs_fs/rep-cache.c

(svn_fs_fs__get_rep_reference): If the database isn't open, just return a

NULL rep, meaning we didn't find a match.

(svn_fs_fs__inc_rep_reference, svn_fs_fs__set_rep_reference): If the database

isn't open, just no op.

* subversion/libsvn_fs_fs/rep-cache.h

(svn_fs_fs__set_rep_reference, svn_fs_fs__get_rep_reference,

svn_fs_fs__inc_rep_reuse): Document new behaviors.

* subversion/libsvn_subr/auth.c

(svn_auth_get_platform_specific_provider): Use proper syntax when checking for

GNOME Keyring and KWallet services. Follow-up to r34085.

Found by: hwright

* notes/http-protocol-v2.txt: rename 'revprop stub' to 'revision stub'
Refactor how Subversion accesses platform-specific auth providers.

* build.conf

(auth-test): New section for the auth tests.

(__LIBS__): Added auth-test to lib.

* subversion/libsvn_subr/cmdline.c

(get_auth_provider): Removed.

(svn_cmdline_create_auth_baton): Use the new API to access

platform-specific auth providers.

* subversion/libsvn_subr/auth.c

(svn_auth_get_platform_specific_provider): Added.

* subversion/tests/libsvn_subr

(svn:ignore): Ignore auth-test.

* subversion/tests/libsvn_subr/auth-test.c: Added.

* subversion/bindings/javahl/native/SVNClient.cpp

(get_auth_provider): Removed.

(SVNClient::getContext): Use the new API to access platform-specific

auth providers.

* subversion/bindings/ctypes-python/

(build.get_apr_config): Removed the libsvn_auth_gnome_keyring and

libsvn_auth_kwallet libs.

* subversion/bindings/swig/core.i,

* subversion/bindings/swig/svn_client.i: Add explicit ignores for all

platform-specific auth providers.

* subversion/bindings/swig/python/tests/

(test_conditional_auth_provider_support): Removed.

* subversion/bindings/swig/include/svn_global.swg


* subversion/include/svn_auth.h

(svn_auth_get_platform_specific_provider): Added.










svn_auth_get_kwallet_ssl_client_cert_pw_provider): Removed exposure to

swig and ctypesgen.

    • ?
Correct a bug in the WC entries walker. In "show_hidden" mode, it tried to

recurse into a 'deleted' or 'absent' directory.

There is no regression test. I tested by modifying svn_client_info() to use

"show_hidden=TRUE" in the call to svn_wc_walk_entries3(), and modifying to run "svn info -R <target>" before resolving. It

failed on test 13.

* subversion/libsvn_wc/entries.c

(walker_helper): Do not try to recurse into a 'deleted' or 'absent'