[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: 1.6.3 up for signing/testing

From: Daniel Shahaf <d.s_at_daniel.shahaf.name>
Date: Sat, 20 Jun 2009 10:20:13 +0300 (Jerusalem Daylight Time)

Jens Peters wrote on Sat, 20 Jun 2009 at 08:43 +0200:
> Tested:
> [file|svn|https(serf)|https(neon)] x [file(default)|[file (sharding and
> packing)|bdb]
> [javahl] x [file|bdb]
> python bindings
> perl bindings
>
> Failures:
> serf tests:
> switch_tests.py 19: out of date propset should fail after a relocate
>
> file (sharding and packing, win-tests.py --fsfs-sharding 2
> --fsfs-packing) tests:
> commit_tests.py 15: hook testing
> commit_tests.py 36: post commit hook failure case testing
> (http://svn.haxx.se/dev/archive-2009-05/0153.shtml)
>

Didn't see that email until now; thanks for raising it again.

When you use --fsfs-{sharding,packing}, the packing is done by
installing a post-commit hook:

    ... from svntest/main.py ...
    # post-commit
    # Note that some tests (currently only commit_tests) create their own
    # post-commit hooks, which would override this one. :-(
    if fsfs_packing:
      # some tests chdir.
      abs_path = os.path.abspath(path)
      create_python_hook_script(get_post_commit_hook_path(abs_path),
          "import subprocess\n"
          "import sys\n"
          "command = %s\n"
          "sys.exit(subprocess.Popen(command).wait())\n"
          % repr([svnadmin_binary, 'pack', abs_path]))
    ... ...

If I remember correctly, commit_tests 15 36 are the only two tests that
install their own post-commit hook. If that's true, it would be
reasonable to assume that what you observed was indeed a "glitch" caused
by one of the post-commit hooks replacing the other.

One way to test this is to build with -DPACK_AFTER_EVERY_COMMIT
-DSVN_FS_FS_DEFAULT_MAX_FILES_PER_DIR=2, run the test suite *normally*
(without $FSFS_SHARDING and without --fsfs-sharding), and see if the
failure persists.

(On the other hand, it is currently *expected* that svnadmin_tests 12 or
17 should FAIL when packing is enabled *at compile-time* --- unless
someone fixed it when I wasn't looking. I should document all these
details somewhere...)

Thanks for testing,

Daniel

> python bindings:
> test_status_editor (wc.SubversionWorkingCopyTestCase)
> (http://svn.haxx.se/dev/archive-2009-05/0087.shtml)
>
>
> Vista SP1 x64, VC 2008,
> apr 1.3.5, apr-util 1.3.7, apr-iconv 1.2.1, httpd 2.2.11, openssl
> 0.9.8k, bdb 4.7.25, zlib 1.2.3, sasl 2.1.23, neon 0.28.4, serf 0.3,
> sqlite 3.6.14
> (everything compiled for x64)
>
> ------------------------------------------------------
> http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2363749
>

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2363751
Received on 2009-06-20 09:20:40 CEST

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.