The {posargs} are global for the current tox process, and we already use them to pass additional arguments to the pytest command in the default testenv, so we can't at the same time also use them to pass options to coverage.py.
The 'htmlcov' tox environment produces the combined coverage.py report, and store it in htmlcov/index.html; this is listed in the tox.ini 'envlist', so it is run by default at every 'tox' invocation, when no explicit env list is passed via either -e option or $TOXENV variable.
The options to `coverage html` are not much interesting anyway; it's much more useful being able to, e.g., run tox for specific tests matching a string or expression:
$ tox -- -k "feaLib"
When running tox locally during development, there's no need to also test python 3.5.
2.7 and 3.6 are enough, and the CI takes care of the others (3.4, 3.5, pypy, etc.).
The --pyargs option makes pytest try interpreting arguments as importable python package names, or if that fails treat them as normal filesystem paths.
Now running `pytest` command will automatically run all the tests in "Tests" folder, as well as the embedded doctests in the installed "fontTools" package.
This is in reply to a comment by Behdad in
fdbf066a8c
Gosh! This is the kind of thinko that doesn't happen with less dynamic languages...
Good one hour of debugging at TYPO Labs. Glad this one is fixed finally!
Fixes https://github.com/fonttools/fonttools/issues/888
Tolerance is set to .5 by default. Most designspaces never hit this since
for most, all deltas are integers anyway. Moreover, all such detas will
round to zero anyway.
This reverts commit d83c2fb2090fa58f94fdf32d1aa2d2d8665cec05.
This is wrong. We should insert if and only iff there's at least one of the
fonts with a Format1. Yours doesn't look at all fonts. Since this one is not a
bugfix but improvement, I'll fix it separately.
Class2Record is mutable, hence doing `[rec2] * l.Class2Count` produces a list containing multiple references to the same instance. When later on we do the interpolation, modifying one would modify them all.
We need to instantiate distinct objects.
See 6f41c2ab53 (commitcomment-21310803)
