subversion

Checkout Tools
  • last updated 4 hours ago
Constraints
Constraints: committers
 
Constraints: files
Constraints: dates

Changeset 1390054 is being indexed.

Sync the 'master-passphrase' branch with recent trunk changes.

(Merged /subversion/trunk:r1383955-1390050.)

  1. /branches/master-passphrase/tools/server-side
  2. … 53 more files in changeset.
On the 10Gb branch: nicer formatting of the numbers produced by svn-bench.

Also, fix "svn-bench null-log -g".

* subversion/include/private/svn_string_private.h

(svn__ui64toa_sep,

svn__i64toa_sep): declare new private utility API

* subversion/libsvn_subr/string.c

(ui64toa_sep): core function

(svn__ui64toa_sep,

svn__i64toa_sep): implement new API

* tools/client-side/svn-bench/main.c

(svn_cl__option): forgot to add support for the "-g" option

* tools/client-side/svn-bench/null-export-cmd.c

(svn_cl__null_export): use nicer output formatting

* tools/client-side/svn-bench/null-list-cmd.c

(svn_cl__null_list): ditto

* tools/client-side/svn-bench/null-log-cmd.c

(log_entry_receiver): fix counters for merged info

(svn_cl__null_log): use nicer and conditional output formatting

    • ?
    /branches/10Gb/subversion/libsvn_subr/string.c
* subversion/tests/cmdline/stat_tests.py

(status_unversioned_dir): Create a nested unversioned directory inside the

standard test working copy, to avoid an unexpected error message when this

test is run within a working copy with an unsupported newer format.

The 'node not found' error is now the only possible error condition

so stop checking for 'not a working copy'. We'll need a separate test

for the latter error condition -- this test would somehow have to avoid

failing within a working copy using an unsupported format.

Suggested by: philip

* subversion/tests/cmdline/stat_tests.py

(status_unversioned_dir): Don't use 'dir' as variable name, since it is

a special term in python (my editor highlights it). Call it 'path'

instead. Also, pass the path to 'svn status' just once, not twice.

* publish/docs/release-notes/1.8.html: The new automatic merge feature seems

to be happy with local modifications in the working copy.

Change a statement I added which was claiming the contrary.

* STATUS: Nominate r1389878.

* STATUS: Propose r1387943, vote for r1389499.
On the 1.7.x-r1389878 branch, merge r1389878 from trunk, resolving conflicts.

Create backport branch for r1389878.

Fix working copy access via a symlink which points to a working copy nested

within an unrelated working copy of an unsupported format.

This patch fixes special_tests 19 and 20 on the 1.7.x branch when run

within a format 30 working copy (backport nomination will follow).

To reproduce the problem fixed by this commit with just a trunk client:

- Check out a working copy WC1.

- Check out one nested working copy WC2 within WC1's root dir.

- Use sqlite3 to set "pragma user_version = 31" in WC1's .svn/wc.db.

- Create a symlink S within WC1 which points at WC2.

- Try to run 'svn status' on the symlink S.

* subversion/libsvn_wc/wc_db_wcroot.c

(svn_wc__db_wcroot_parse_local_abspath): If we find an unsupported working

copy after traversing upwards from a symlink, and the symlink points to

a directory, try to find a working copy at that directory instead.

On the 10Gb branch: minimize client-side processing overhead for

null-export in quiet mode. For ra_svn, skip through the command

stream without further processing as to not become the bottleneck

when running against a fast server. The other RA modules don't

have that problem, yet.

* subversion/libsvn_ra_svn/ra_svn.h

(svn_ra_svn__read_command_only): declare new private API

* subversion/libsvn_ra_svn/marshal.c

(readbuf_skip,

read_command_only): new skip-command processing utilities

(svn_ra_svn__read_command_only): implement new private API

* subversion/libsvn_ra_svn/editorp.c

(svn_ra_svn_drive_editor2): allow for EDITOR to be NULL;

skip command processing if there is no editor

* tools/client-side/svn-bench/null-export-cmd.c

(bench_null_export): add QUIET parameter; don't use an

editor for ra_svn in that mode

(svn_cl__null_export): update

* STATUS: Nominate r1389499 and r1389851.

On the 1.7.x-r1389851 branch, merge r1389851 from trunk, resolving conflicts.

Create a backport branch for r1389851.

Fix commit from multiple working copies which are nested within an unrelated

working copy.

This patch fixes commit_test 26 on the 1.7.x branch when run within

a format 30 working copy (backport nomination will follow).

To reproduce the problem fixed by this commit with just a trunk client:

- Check out a working copy WC1.

- Check out two nested working copies within WC1's root dir.

- Make two non-overlapping changes, one within each nested working copy.

- Use sqlite3 to set "pragma user_version = 31" in WC1's .svn/wc.db.

- Try to commit the nested working copies simultaneously by listing them

as targets for 'svn commit'.

* subversion/libsvn_client/commit.c

(svn_client_commit6): When committing from multiple working copies, get the

RA editor from the first working copy, rather than using the pre-computed

common base path of commit targets. The base path might point into an

unrelated parent working copy of nested working copies, which might point

us to the wrong repository or could cause an error if we cannot handle the

parent's working copy's format.

Merge r1378847 from trunk:

* r1378847

Make GNOME keyring library work with very old glib.

Justification:

Keyring library fails to load at runtime and it's hard for the user

to work out why.

Votes:

+1: philip, brane, stsp

+0: danielsh

On the 10Gb branch: Derive a benchmark client from our standard CL client.

Instead of actual I/O, only collect statistics. In quiet mode, reduce

client-side processing to just handling the network communication (to

the degree possible).

Currently, only null-export, null-list and null-log are being provided.

Others can be added easily by copying from the respective command from

the standard client and adapting it to benchmarking needs.

* build.conf

(svn-bench): new tool

(libs): add tool reference

* tools/client-side/svn-bench

new tool folder

* tools/client-side/svn-bench/cl.h

copy of subversion/svn/cl.h; minimize it; adapt to null-* commands

* tools/client-side/svn-bench/util.c

ditto

* tools/client-side/svn-bench/main.c

ditto

* tools/client-side/svn-bench/notify.c

ditto

* tools/client-side/svn-bench/help-cmd.c

(svn_cl__help): adapt UI texts

* tools/client-side/svn-bench/null-export-cmd.c

copy of subversion/svn/export-cmd.h; implement minimal editor that

counts incoming data

* tools/client-side/svn-bench/null-list-cmd.c

copy of subversion/svn/list-cmd.c; replace output with simple counters

* tools/client-side/svn-bench/null-log-cmd.c

ditto

  1. /branches/10Gb/tools/client-side/svn-bench
Fix the build of the Perl bindings with newer versions of SWIG.

* subversion/bindings/swig/perl/native/Makefile.PL.in

(cflags): Remove -ansi from the CFLAGS that we're using since SWIG is not

producing ANSI compliant code.

Suggested by: philip

On the auto-props-sdc branch: Sync with

^/subversion/branches/inheritable-props through r1389501.

  1. /branches/auto-props-sdc/tools/server-side
    • ?
    /branches/auto-props-sdc/subversion/svn/cl.h
    • ?
    /branches/auto-props-sdc/build.conf
  2. … 67 more files in changeset.
Treat unsupported WC formats as "unknown" rather than raising an error.

Many tests currently fail on the 1.7.x branch if run within a format 30

(i.e. trunk client) working copy. The most prominent failure is an

error from 'svn checkout' which refuses to create a format 29 working

copy within a format 30 one, even though the two working copies are

suposed to be entirely unrelated.

If this patch is applied to 1.7.x, most tests now pass in a format 30

working copy, except for the following:

FAIL: commit_tests.py 26: commit from two working copies

FAIL: special_tests.py 19: operate on symlink to wc

FAIL: special_tests.py 20: svnversion on symlink to wc

FAIL: stat_tests.py 19: status on unversioned dir

* subversion/libsvn_wc/lock.c

(svn_wc__internal_check_wc): If the format isn't supported, just return

a format number of zero, rather than raising an error indicating that

the working copy is not supported.

On the inheritable-props branch: Remove the branch TODO re server dictated

config with iprops, that is being looked at on another branch.

* BRANCH-README: Update

On the auto-props-sdc branch: Update the BRANCH-README.

    • ?
    /branches/auto-props-sdc/BRANCH-README
On the inheritable-props branch: Complete the bump to WC format 31.

* subversion/libsvn_wc/wc.h:

(SVN_WC__VERSION): Oops, forgot to actually bump the format in r1389446.

On the inheritable-props branch: Sync with ^/subversion/trunk through

r1389356.

Resolve some minor conflicts with the incoming bump to WC format 30 (see

r1387742). We had already made that bump on this branch to support a schema

change required for iprops (see r1376154 and r1376924). With this sync, the

latter change is now part of the bump to format 31.

  1. … 43 more files in changeset.
* tools/server-side/fsfs-reorg.c

(get_content): fix another issue with 32 bit file pointers

* STATUS: Vote for r1378847, approving.

* STATUS: Use non-ambiguous syntax to show partial votes for r1374800 group.

* STATUS: Add r1389364 to r1374800 group and vote.

Add a regression test for problems fixed in r1374800 and r1374802.

* subversion/tests/cmdline/patch_tests.py

(patch_git_with_index_line, test_list): New test.

Make the FSFS reorg tool independent of apr_off_t's size.

Always use 64 bits.

* tools/server-side/fsfs-reorg.c

s/apr_off_t/apr_int64_t/g

Found by: stsp