Closed tony closed 6 years ago
Workaround is to use --skip-lock
for now
Oh huh totally forgot about that, was not intending to just sit on it 😦 Will cut 1.3.2. Thanks for the report!
1.3.2 should be on PyPI now, despite some irritating mishaps. Please follow up if that doesn't install better under your setup.
Hrm yup today's just not a good day, Sphinx 1.6 seems to be breaking us pretty badly. Look out for a 1.3.3...
Suspect the real issue is the pin itself was wrong, there is no sphinx 2.x as far as I can tell, but it's either 1.5 or 1.6 that started breaking their internal APIs which in turn broke us. I may need to rephrase it as sphinx>=1.3,<1.6
? Checking.
Yea 1.6 changed arity of app.config.init_values
. Put in a workaround; now my test suite passes, though we still get a grazillion ANSI-red warnings, which I will try to fix next lest I get bombarded with issues for that 😆
Those warnings seem to have been in since 1.4, so I bet this is actually related to the arity thing, since...the arg that came and then went, was a warnings runner! which we were handing a dummy/noop lambda into - thus muting those warnings.
Gonna see if there's a better way, such as configuring the app at runtime, since there ought to be specific config options for these specific warnings (going by their issues...)
Grump, got the unit test suite's view of things squared away, only to find some more Sphinx API changes when I ran the integration suite. Still digging. Also adding a full test matrix to Travis so this sort of thing doesn't cause regressions.
Cool. I think that has all been ironed out, and the matrix is happy, meaning hopefully we didn't torch any existing setups. Took the time to drop 2.6/3.3 support (those cells were dying thanks to Sphinx anyway) and it will be out as Releases 1.4.0 momentarily.
https://github.com/bitprophet/releases/compare/1.3.1...master
f5fcad3 didn't make it in