On Fri, Jul 03, 2020 at 10:11:21PM +0100, Laurence Tratt wrote:
> I should have put NO_TEST=Yes here. The way the test suite is run is fairly
> idiosyncratic and would require more patching than I think we want to
> maintain. For example, by not following the documentation, I can make most
> of the tests run, but at one point something seems to require that tqdm is
> installed, which it isn't, and then a couple of tests fail.
Sounds reasonable to me, but I'd put a comment above the NO_TEST line
briefly mentioning the extra work as reasons for disabled tests.
You could simply install/require the latest tqdm version as follows,
would that help?
TEST_DEPENDS = ${PKGPATH}=${MODPY_EGG_VERSION}
I also quickly installed flake8, py3-nose and py3-coverage as TDEPs and
`make test' does a lot of tests: some SKIP, some ok, some ERROR, but it
seems stuck at the following test:
Test concurrent.futures.ThreadPoolExecutor ...
I had to kill the python3.7 process, ^C wouldn't stop it.
> I've removed MODPY_EGG_VERSION but I thought it was preferred to use
> MASTER_SITES over GH_*? I might be very wrong on that.
That's the case for tarballs under *stable* /release/ URLs, but with
auto-generated tarballs from /archive/ we use GH_*.
Another benefit of using GH_* (and no explicit MASTER_SITES, DISTNAME,
PKGNAME or WRKDIST) is having a proper distinfo, yours currently just
records the version but not the project name in the file name.
