tigefa4u projects Trac

0.11-stable: Fixed the left margin of the wikitoolbar so that it aligns with its associated `<textarea>`.

Also fixed the positioning and sizing of the wiki page editor `<textarea>`.

rework-testing: testing with MySQL as a backend
- slight simplification of the Popen call (remove arguments with default values, no need for PIPE)
- make sure the call works on Windows

Hopefully this still works on Unix...

rework-testing: the Makefile can now be used on Windows (bash used as the shell for the Makefile, though)

rework-testing: forgot to check for svn:eol-style...

Merge current state of rework-testing on 0.11-stable (backport of relevant parts of r8131)

0.12dev: merged rework-testing branch [8000:8130/sandbox/rework-testing]

rework-testing branch: the "empty" result value from `FunctionalTestEnvironment.repo_path_for_initenv` was problematic with Python 2.4's subprocess, at least on Windows.

{{{
Initenv for '...\testenv\trac' failed.
Wrong number of arguments: 3

Traceback (most recent call last):
...
File "...\trac\tests\functional\testenv.py", line 189, in _tracadmin
raise Exception('Failed with exitcode %s running trac-admin ' \
Exception: Failed with exitcode 2 running trac-admin with ('initenv', 'testenv8324', 'sqlite:db/trac.db', 'svn', '')
}}}

Returning `''` fixes it, and doesn't appear to be problematic for higher versions of subprocess (2.5 and 2.6).

rework-testing branch: fix `FunctionalTestEnvironment.destroy` on Windows, which wasn't able to remove the test environment due to the newly introduced get_trac_environment call:
- the db/trac.db was still opened for reading
- the log/trac.log was still opened for writing

Doing a `env.shutdown()` was enough to get rid of the first problem, but for the second, the log handler had also to be cleaned-up, so shutdown was modified to that effect.

There's apparently still one user of the logger ''after'' the environment has been destroyed, as we see the following:
{{{
...
Test attachment: raw format as explicit argument ... ok
No handlers could be found for logger "...\repos\rework-testing\testenv\trac"
Test for regression of the plugin reload fix in r6017 ... ok
...
}}}
but so far I couldn't locate it (not ''that'' important either).

Follow-up to r8080, which broke Python 2.3 support.

In Python 2.3 on Windows (and apparently also on the Darwin 9.6.0 platform used for the buildbot), the file objects associated to the handler's connection socket can't be queried for their "closed" status, as they always returns "True").

0.12dev: Merged [8126] from 0.11-stable.