"Bert Huijben" <[email protected]> writes: >> > It might also be possible to write an annotation for the test that >> > configures the environment to enable the sleep within the libraries >> > for that particular test. It would be a bit less granular (on the >> > order of a test, not a specific command), but would be easier to >> > maintain once the initial plumbing is in place.
Something like this would be good. Note that we don't want to change SVN_I_LOVE_CORRUPTED_WORKING_COPIES_SO_DISABLE_SLEEP_FOR_TIMESTAMPS for the python process as this affects all the tests running in parallel. What we really want is a way to control the python process to clear the env var for a single svn invocation. Note that at present some tests do use do_sleep_for_timestamps explicitly and this has the side-effect of changing the environment for all parallell tests. The tests that use do_sleep_for_timestamps don't use no_sleep_for_timestamps because it is not possible for a test to do that without potentially breaking a parallel test. >> Index: subversion/tests/cmdline/svntest/actions.py >> ========================================================== >> ========= >> --- subversion/tests/cmdline/svntest/actions.py (revision 1207768) >> +++ subversion/tests/cmdline/svntest/actions.py (working copy) >> @@ -1180,7 +1180,9 @@ def run_and_verify_patch(dir, patch_path, >> >> # Update and make a tree of the output. >> patch_command = patch_command + args >> + do_sleep_for_timestamps() >> exit_code, out, err = main.run_svn(True, *patch_command) >> + no_sleep_for_timestamps() No, you can't use no_sleep_for_timestamps if tests are running in parallel. >> Index: subversion/libsvn_client/patch.c >> ========================================================== >> ========= >> --- subversion/libsvn_client/patch.c (revision 1207768) >> +++ subversion/libsvn_client/patch.c (working copy) >> @@ -2926,6 +2926,8 @@ apply_patches(/* The path to the patch file. */ >> SVN_ERR(svn_diff_close_patch_file(patch_file, iterpool)); >> svn_pool_destroy(iterpool); >> >> + svn_io_sleep_for_timestamps(abs_wc_path, scratch_pool); >> + > > It's not the patch code that causes the problem. The operation before the > patch should have waited until it is safe to exit. Exactly. -- uberSVN: Apache Subversion Made Easy http://www.uberSVN.com
