Btest

* Separating semantics of groups and thread serialization into
separate options. -g still specifices TEST-GROUPs that are to be
executed, but these groups don't any longer control which tests
get serialized in a parallel execution. For that, there's a new
"TEST-SERIALIZE: <tag>" command that takes a tag and then makes
sure all tests with the same tag are run within the same thread.
(Robin Sommer)

* TEST-GROUPS can now be given multiple times now to assign a test
to a set of groups. (Robin Sommer)

* Extended -g to accept a comma-separated list of groups names to
run more than one test group. (Robin Sommer)

* New output handler for console output. This output is now the
default when stdout is a terminal. It prints out a compressed
output that updates as btest goes through; it also indicates the
progress so far. If btest's output is redirected to a
non-terminal, is switches back to the old style. (Robin Sommer)

* New test command TEST-NO-FILTER: <filter>

This allows to ignore a test when running a specific filter. (Robin Sommer)

* Changing the way filters are activated.

-F <filter> now activates only the given filter, but doesn't run
the standard tests in addition. But one can now give -F a
command-separated list of filters to activate them all, and refer
to the standard tests without filter as ``-``. (Robin Sommer)

* Fix to allow numbered test to be given individually on the command
line. (E.g., integer.geq-3 for a file that contains three tests).
(Robin Sommer)

* A number of smaller fixes for bugs, plus polishing, caused by the
recent restructuring. (Robin Sommer)

* Removing the error given when using -r with tests on the command
line. It's unnessary and confusing compared to when listing tests
in btest.cfg. (Robin Sommer)

* Adding a new "finalizer" option.

``Finalizer``
An executable that will be executed each time any test has
succesfully run. It runs in the same directory as the test itself
and receives the name of the test as its parameter. The return
value indicates whether the test should indeed be considered
succeeded. By default, there's no finalizer set. (Robin Sommer)

* btest is now again overwriting old diag files instead of appending
(i.e., back to as it used to be). (Robin Sommer)

* Diag output is now line-buffered. (Daniel Thayer)

* Adding new option -r that reruns all tests that failed last time.
btest now always records all failed tests in a file called. (Robin
Sommer)

* Internal restructuring to factor output out into sublcasses.
(Robin Sommer)

* Adding parallel test execution to btest. (Robin Sommer)

- A new option "-j <n>" allows to run up to <n> tests in
parallel.

- A new TEST-GROUP directive allows to group tests that can't
be parallelized. All tests of the same group will be
executed sequentially.

- A new option "-g <group>" allows to run only tests of a
certain group, or with "-g -" all tests that don't have a
group.

* New environment variable TEST_DIFF_BRIEF. If set btest-diff no
longer includes a mismatching file's full content it the
diagnostic output. This can be useful if the file being compared
is very large. (Robin Sommer)

* Fix btest-bg-wait's kill trap and -k option. (Jon Siwek)

* A new option -u for interactively updating baselines.

* Teach btest's TEST-START-FILE to make subdirectories (Jon Siwek)

* Output polishing. (Robin Sommer)

* Have distutils install 'btest-setsid' script. (Jon Siwek)

* A portable setsid. (Robin Sommer)

* Fixes for background execution of processes.

* Fixing exit codes. (Robin Sommer)