subversion

Checkout Tools
  • last updated 2 hours ago
Constraints
Constraints: committers
 
Constraints: files
Constraints: dates

Changeset 1767153 is being indexed.

Finalize the svn_repos_list interface: Name scratch pools explicitly.

* subversion/include/svn_repos.h

(svn_repos_dirent_receiver_t,

svn_repos_list): The pools given here are pure SCRATCH_POOLs.

* subversion/libsvn_repos/list.c

(fill_dirent,

report_dirent,

do_list,

svn_repos_list): Same.

Make the new svn_repos_list easier to use with typical client requests.

Instead of a single pattern, accept an array of alternative patterns of

which at least one must match if any patterns are given. Sort the output

by path to make it match the client-side implementation's output order.

* subversion/include/svn_repos.h

(svn_repos_list): We now support multiple glob patterns and sort the

output.

* subversion/libsvn_repos/list.c

(matches_any): New utility function for matching against an array

instead of a single pattern.

(report_dirent): For efficiency reasons, we now filter early before

calling this reporter function.

(filtered_dirent_t,

compare_filtered_dirent): New filtering and sorting support.

(do_list): Support multiple PATTERNS. Filter and sort the data before

processing it further.

(svn_repos_list): Update signature, filtering and passing of parameters.

* subversion/tests/libsvn_repos/repos-test.c

(test_list): Adapt test case to interface change.

Fix a copy-n-pasto in the new svn_repos_list code.

Found by: danielsh

* subversion/libsvn_repos/list.c

(svn_repos_list): The DEPTH parameter is relevant for the top-level dir.

* site/publish/docs/community-guide/general.part.html: Fix links to hackathon

news items on elego's web site and add the link for 2016's hackathon.

Reminded by: danielsh

* STATUS: Nomniate r1766699, r1766704, r1766711.

Avoid accessing a non-existent path while preparing a reintegrate merge.

* subversion/libsvn_client/merge.c

(find_unsynced_ranges): Reparent the RA session to TARGET_LOC only if

this location actually exists within the unmerged revision range.

* subversion/tests/cmdline/merge_reintegrate_tests.py

(reintegrate_noop_branch_into_renamed_branch): Remove XFAIL marker.

Rename recently added test and fix its expectations.

* subversion/tests/cmdline/merge_reintegrate_tests.py

(reintegrate_unsynced_into_renamed_branch): Rename to ...

(reintegrate_noop_branch_into_renamed_branch): ... this. Explain the test

setup correctly, and expect the final reintegrate merge to succeed.

(test_list): Track renamed test.

Add a regression test which shows that reintegrate merges have a problem

if the merge target branch was renamed: It tries to access the new

(post-move) path at a revision where it didn't exist yet.

* subversion/tests/cmdline/merge_reintegrate_tests.py

(reintegrate_unsynced_into_renamed_branch, test_list): New test.

Merge r1766240 from trunk:

* r1766240

Fix a test failure in Ruby bindings due to modifying a frozen object

(NilClass)

Justification:

Fixes ruby test failures with ruby >= 2.2

Votes:

+1: jamessan

+0: rhuijben

* STATUS:

Add r1766352 group (issue #4658).

On the 1.8.x-issues4658 branch:

Merge r1766352 and r1766590 from /trunk, resolve tree due to

refactoring done in 1.9 and adapt to a couple of renames.

    • ?
    /branches/1.8.x-issue4658/subversion/libsvn_fs_fs/fs_fs.c
Open backport branch for the issue #4658 group.
* STATUS:

Add r1766352 group (issue #4658).

Add a test case for issue #4658.

* subversion/tests/libsvn_fs_fs/fs-fs-pack-test.c

(large_delta_against_plain): New test.

(test_funcs): Register new test.

* subversion/tests/libsvn_subr: Add 'xml-test' to ignore list.

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

(test_invalid_xml_signal_bailout): Fix typo in test error message.

Fix a spurious test failure on OpenBSD due to a bug in the test:

subversion/tests/libsvn_subr/xml-test.c:184: (apr_err=SVN_ERR_TEST_FAILED)

svn_tests: E200006: Got unxpected error '(null)'

FAIL: xml-test 4: test svn_xml_signal_bailout() for invalid XML

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

(test_invalid_xml_signal_bailout): SVN_TEST_ASSERT_ANY_ERROR() will clear the

error so make a copy of its status value and use it instead of err->apr_err.

Found by: bb-openbsd buildbot

* subversion/libsvn_client/conflicts.c

(diff_file_added): Fix file existence check. It was testing the same

variable twice but should check two different ones instead.

Found by: jamessan

(with Coverity)

* tools/dist/advisory.py (top-level, parse_args): Fix typo in script description.

* subversion/tests/cmdline/depth_tests.py

(fold_tree_with_deleted_moved_items): Replace two tabs with equivalent

spaces.

(test_list): Replace tab with spaces and re-indent the closing ']' to

the same level as used in other test files.

* STATUS: Remove unneeded comment. Answered on irc
* STATUS: Approve ruby test fix.
* STATUS: Cast some gpg support votes after checking gpg sources
* STATUS: Nominate r1766327.
* STATUS: Nominate r1766323.
[Issue #4658]

Fix a long-standing FSFS bug that causes read errors on valid / non-corrupted

representations. The problem occurs if

* the representation is a DELTA against a PLAIN one,

* spans multiple delta windows, and

* has a matching sub-string in some window after there were no matches

in the previous window.

Because we used to read the PLAIN base window only if it was really needed

due to a matching sub-string In that case, the read offset inside the PLAIN

rep would not be updated. As a result, when the next window needs to be

read as the delta base, the previous one is read instead. This would almost

certainly result in a corrupted reconstructed fulltext. The read will

ultimately fail due to a checksum mismatch.

A test case and issue tracker entry will follow.

* subversion/libsvn_fs_fs/cached_data.c

(skip_plain_window): New function.

(get_combined_window): Always move the offset within the base rep in sync

with the delta rep.

Improve discovery of gpg-agent sockets.

In r1600368, support was added for discovering sockets of running

gpg-agents. However, this was only checking $GPG_AGENT_INFO or

$HOME/.gnupg/S.gpg-agent. $GNUPGHOME may also be set in the environment

to use a location other than $HOME/.gnupg as gpg's home directory.

* subversion/libsvn_subr/gpg_agent.c

(find_running_gpg_agent): Use $GNUPGHOME/S.gpg-agent as the first fallback

when $GPG_AGENT_INFO is not present in the environment. Clarify a comment

about the meaning of "home directory" in terms of gpg.

* subversion/libsvn_subr/gpg_agent.c

(simple_gpg_agent_next_creds): Unconditionally call bye_gpg_agent when we're

done with gpg-agent to fix a file descriptor leak.

* STATUS: Nominate r1766240.
Fix a test failure in Ruby bindings due to modifying a frozen object (NilClass)

Starting with Ruby 2.2, nil is a "frozen" object. The test_load test calls

repos.load_fs(nil), which leads to nil being used to create the dumpstream

in svn_swig_rb_make_stream.

* subversion/bindings/swig/ruby/test/test_repos.rb

(test_load): Pass a new, transient Object into repos.load_fs