In addition to the conversion to Python, run_all now defaults to running
all Python tests in parallel, using the same approach previously
available via '--fast'. If desired one can force all tests to run
serially by calling run_all with --serial
A debugging flag was now also included in run_all. Pass one or more -l
(-l, -ll or -lll) for different levels of debugging information.
The problems script now outputs color on each of the test categories,
following the same rules used by simpletap.
It also now includes a --summary switch which outputs the same short
report seen when using ./run_all.
- Relocated (relevant) tests from bug.concat.t to uda_orphan.t, which yields
better grouping, and allows the deletion of bug.concat.t, which contains
tests that are no longer relevant.
- This is a large commit, as all the changes are centered around the elimination
of CLI.
- CLI is no longer compiled.
- Context no longer maintains CLI + CLI2.
- Filter now walks the parse tree and sends to Eval a
std::vector <std::pair <std::string, Lexer::Type>> containing only args tagged
with FILTER.
- Filter more efficiently sets/unsets Eval::debug, by doing it less often.
- The filterExpr.length() check is no longer meaningful, and instead the size of
the std::vector above is used.
- Filter::pendingOnly performs better analysis.
- Filter::safety makes use of the std::vector size also.
- Task::modify makes use of 'canonical' rather than 'name', which is a policy
change, not a fix.
- If an argument is an exact match for a command ('projects'), it is a command.
- If an argument is not an exact match for command ('project'), but is an exact
match for an attribute, it is not a command.
- If an argument fails the above tests, but canonicalizes to a command, it is a
command.
- Make cmake complain and abort the build process if libgnutls is not
available. In order to build Taskwarrior without "sync" support,
or to build it on a system with libgnutls missing, the new cmake
flag "-DENABLE_SYNC=OFF" can be used.
