subversion

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

Changeset 878115 is being indexed.

* COMMITTERS: Definitively mark the end of the full committers

section, to solve the "mailing list membership script breaks

unpredictably when we make otherwise reasonable changes"

problem once and for all.

* COMMITTERS: oops, rassilon and kon were partial committers.
* COMMITTERS: bring 'dormant' full-committers list up to date; put it in a more visible place, right after the active ones.
* subversion/libsvn_subr/stream.c

(read_handler_apr): Fix typo in comment.

Follow-up to r38036:

* subversion/include/svn_io.h

(svn_io_line_filter_cb_t): Fix typo in argument list.

Allow filtering while reading lines from a stream with svn_stream_readline().

* subversion/libsvn_subr/stream.c

(svn_stream_t): Add line_filter_cb member.

(svn_stream_create): Initialise new member.

(svn_stream_set_line_filter_callback): New function, allows setting a

line-filtering callback on a stream.

(line_filter): New helper function for svn_stream_readline() to call

the stream's line-filtering callback, if defined.

(svn_stream_readline): Only return lines if the line filtering callback

did not filter them. If the end of the stream is reached and the last

line is also filtered, return an empty stringbuf.

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

(test_stream_line_filter): New test to test line-filtering on streams.

(line_filter): Line-filering callback for the test_stream_line_filter test.

(test_funcs): Add test_stream_line_filter.

* subversion/include/svn_io.h

(svn_io_line_filter_cb_t): New callback type.

(svn_stream_set_line_filter_callback): Declare.

(svn_stream_readline): Update docstring to describe line-filtering.

Follow-up to r38025:

Fix stream-test 3 ("test streams reading from range of file") on 64-bit

architectures.

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

(test_stream_range): Use 'apr_size_t' type for 'len' variable.

* subversion/libsvn_diff/parse-diff.c

(svn_diff__parse_next_patch): Do not assume that filenames in unidiffs

are always terminated by a tab. Just use the tab as a separator if

it is present. GNU patch doesn't bail on patches where filenames aren't

terminated by a tab, so we shouldn't either.

Found by: arfrever

* www/getting.html: Ask people to keep list of operating systems and

package providers in alphabetical order.

Also order Solaris package providers alphabetically.

* subversion/tests/cmdline/commit_tests.py

(commit_multiple_wc_multiple_repos): Fix typo in comment, and tweak

wording of error an message.

Follow-up to r38030:

* subversion/tests/cmdline/commit_tests.py

(commit_multiple_wc_nested): Fix the invocation of run_and_verify_svn()

attempting to commit from multiple working copies.

Pass an error message, and also expect output on stdout and no output

on stderr when expecting a commit to succeed, not the other way around.

Add and tweak some tests for issue #2381.

* subversion/tests/cmdline/commit_tests.py

(commit_multiple_wc): Rename to ...

(commit_multiple_wc_nested): ... this, because it tests the case where

two working copies are nested. Also change this test to expect the

commit to succeed, which is the goal of issue #2381.

(commit_multiple_wc): New test, like commit_multiple_wc_nested, except

that working copies are not nested. Also expects a commit from multiple

working copies to succeed.

(commit_multiple_wc_multiple_repos): New test which tries to commit

from two working copies which do not come from the same repository.

The commit is expected to fail, because commits to multiple repositories

are outside the scope of issue #2381.

(test_list): Track renamed test and add the new tests.

Mark commit_multiple_wc_nested and commit_multiple_wc as XFail,

because issue #2381 is still work in progress.

Patch by: HuiHuang <yellow.flying@yahoo.com.cn>

(tweaked by me)

On the '1.6.x' branch:

* STATUS: Nominate r38004, r38014, r38028.

Follow-up to r38004 and r38014:

* configure.ac: Remove check for setenv().

* subversion/libsvn_auth_kwallet/kwallet.cpp

(INITIALIZE_APPLICATION): Check 'svn:auth:qapplication-safe' parameter

instead of 'SVN_QAPPLICATION_SAFE' environment variable.

(get_wid): Add 'parameters' argument. Check 'svn:auth:qapplication-safe'

parameter instead of 'SVN_QAPPLICATION_SAFE' environment variable.

(get_wallet): Update call to get_wid().

* subversion/svn/main.c

(main):

* subversion/svnsync/main.c

(main): Set 'svn:auth:qapplication-safe' parameter instead of

'SVN_QAPPLICATION_SAFE' environment variable.

Suggested by: rhuijben

On the 'python-3-compatibility' branch:

Merge r37766:38026 from trunk.

  1. … 160 more files in changeset.
* www/getting.html: Add link to OpenCSW under "Solaris".

Patch by: Mike Watters <valholla75@gmail.com>

Tweaked by: me

Implement a stream interface which allows reading inside a range of a file.

'svn patch' will make use of this interface later to read hunks directly

from the patch file, instead of loading them into memory.

Loading hunks into memory can be too expensive e.g. when large files

are added as part of a unidiff.

An additional stream interface which will allow filtering while reading

lines from a stream will also need to be created. This will allow

'svn patch' to make use of streams for the original text, modified text,

and unidiff text of hunks, all backed directly by the patch file.

The amount of memory used by 'svn patch' to manage a hunk will then

be determined by the length of lines inside the hunk, rather than the

size of the entire hunk.

* subversion/libsvn_subr/stream.c

(baton_apr): Add new fields start and end.

(read_handler_apr): If start and end of a range were specified,

do not allow reads outside of the range.

(reset_handler_apr): If the start of a range was specified, reset

the stream to the start of the range instead of the start of the file.

(svn_stream_from_aprfile2): Initialise new start and end fields of

the baton to an invalid range.

(svn_stream_from_aprfile_range_readonly): New function. Allows creating

of a stream to read inside a range of a file.

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

(test_stream_range): New test to test streams reading inside a range of

a file.

(test_funcs): Add new tests.

* subversion/include/svn_io.h

(svn_stream_from_aprfile_range_readonly): Declare.

Follow-up to r38023:

* subversion/tests/cmdline/svntest/testcase.py

(FunctionTestCase.__init__, FunctionTestCase.get_function_name):

Simplify slightly to avoid adding an extra class member variable.

Allow specifying the names of test functions to run.

Saves the hassle of having to run the test file with the --list

option just to find out the index your new shiny test is at in

order to run it.

For example:

$ ./basic_tests.py basic_checkout basic_commit

PASS: basic_tests.py 1: basic checkout of a wc

PASS: basic_tests.py 3: basic commit command

* subversion/tests/cmdline/svntest/testcase.py

(TestCase.get_function_name): New member function in abstract class.

(FunctionTestCase.__init__): Remember name of test function for retrieval

by get_function_name().

(FunctionTestCase.get_function_name): Implement new member function.

* subversion/tests/cmdline/svntest/main.py

(TestRunner.get_function_name): New function, making the private TestCase's

get_function_name() function available to the public.

(usage): Mention the possibility of passing test function names.

(run_tests): When parsing the argument as a number of range of numbers

has failed, try to find a test function to call, which should be named

by the argument.

Recommend r38021 for backport.
Finish issue #3425 - Links generated by ?r=/p= public URL schema

aren't revision-sticky. Per the plan discussion on the dev-list,

we'll canonicalize URLs that carry a working revision (r=REV) --

regardless of whether they also have a peg revision (p=PEGREV) -- into

ones that only carry a peg revision using redirects sent back to the

client.

* subversion/mod_dav_svn/dav_svn.h

(struct dav_resource_private): New 'pegged' member.

* subversion/mod_dav_svn/repos.c

(parse_querystring): Rework this function for readability, to

support the population of the new 'pegged' resource flag, and to

generate "302 Moved Temporarily" for any request in which the node

locations algorithm was invoked.

(get_resource): Minor logic re-org. Also, stop ignoring r->args

when doing a directory-needs-a-trail-slash redirect, because now we

might need to preserve that stuff.

(deliver): If delivering a collection, see if that collection

resource was opened via our public peg-revision CGI query

interface. If so, let the returned dirent links carry the specified

peg revision, too.

* STATUS: Add r38019 to the (r38015, r38017) group, and vote for the

whole lot. Also channel Bert's +1 for r38019 from IRC.

* subversion/tests/cmdline/basic_tests.py

(delete_keep_local_twice): Trim docstring to 50 characters or less

to avoid AssertionError when running basic_tests.py --list.

* STATUS: Nominate r38015,r38017 for backporting
* subversion/libsvn_wc/adm_ops.c

(svn_wc_delete3):

Don't delete unversioned files and directories when keep_local is

true. This fixes the delete_keep_local_twice test added in r38105

for ^/branches/1.6.x.

Remove some unnecessary logic.

* subversion/libsvn_fs_base/revs-txns.c

(svn_fs_base__txn_prop, svn_fs_base__revision_prop): No need to

check for non-NULL-ness of proplist -- the function used to fetch

it promises not to return NULL.

Add new test to show double delete issue in Subversion 1.6.x.

(It appears the WC-NG work fixed this issue)

* tests/cmdline/basic_tests.py

(delete_keep_local_twice): New test to show 1.6.x issue.

(test_list): Add new test.

Follow-up to r38004:

* configure.ac: Add check for setenv() which defines HAVE_SETENV when setenv()

is available.

* subversion/libsvn_auth_kwallet/kwallet.cpp

(get_parent_pid): Simplify.

(get_wid): Rename 'subpool' to 'iterpool' and actually use it.

* subversion/svn/main.c

(main):

* subversion/svnsync/main.c

(main): Check HAVE_SETENV instead of __USE_BSD.

Merge from trunk r37907, whose log message reads thusly:

Make the file-to-stringbuf functions a little smarter about memory usage.

* subversion/libsvn_subr/io.c

(stringbuf_from_aprfile): New helper function, mostly cored from

svn_stringbuf_from_aprfile, but with extra parameters and

functionality for avoiding the stringbuf-realloc loop where possible.

(svn_stringbuf_from_file2): Use stringbuf_from_aprfile() helper function

now, and remove comment about what we should be doing (since now we are).

(svn_stringbuf_from_aprfile): Make this a thin wrapper around the new

stringbuf_from_aprfile() helper file it sired.

Also:

* STATUS

Remove the r37907 grouping.

Merge from trunk r36298, the merge-able portion of which has been

described like so:

* subversion/tests/cmdline/svntest/main.py

(is_ra_type_dav_neon): New function.

(is_ra_type_dav_serf): New function.

Also:

* STATUS

Remove the r36298 group.