You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As I do not have MacOS around to do further experiments, it's unlikely I will be able to solve this by myself. Maybe @hynek or @carlpulley can shed some light?
I use OSX, I've only ever used python-rapidjson on 3.5 and 3.6 and it works perfectly. We don't test py34 with tox. We could add that in there and see if its specific to Py3.4 on OSX
Going to dig a bit more on versions of GCC running here, but I'm just jumping into this, so I don't have an intuition of where to look. If someone can throw me some pointers, I can put some time into it.
I actually just ran into this on my other Mac... Not sure why but I think it might be related to an older version of clang or something. On my mac that I have no issues with:
That's so strange! But system/compiler versions may explain why the trouble does not always manifest itself: for example, the Travis build succeeds except with Python 3.4...
Activity
Disable MacOS wheels
lelit commentedon Sep 11, 2017
As I do not have MacOS around to do further experiments, it's unlikely I will be able to solve this by myself. Maybe @hynek or @carlpulley can shed some light?
hynek commentedon Sep 24, 2017
Relevant errors are:
Could this be some C-header-used-in-C++ weirdness?
Relevant Python code for 3.4 is at:
https://github.com/python/cpython/blob/86a713cb0c110b6798ca7f9e630fc511ee0a4028/Include/pymath.h#L82-L118 and it at least seems to have detected the presence of the functions it's complaining about. So my mostly ignorant guess is that it's less about macOS per se and more about compiler versions/strictness?
sontek commentedon Sep 29, 2017
I use OSX, I've only ever used
python-rapidjsonon 3.5 and 3.6 and it works perfectly. We don't test py34 withtox. We could add that in there and see if its specific to Py3.4 on OSXlelit commentedon Sep 29, 2017
Yes, the problem I got is only with py34.
kenrobbins commentedon Sep 29, 2017
I don't have an easy way to test either, but I think this idea might work except with
math.hinstead ofcmath. 1ff9a18thanosd commentedon Oct 27, 2017
I don't want to hijack the wrong thread, as I'm using Python 3.6, but I'm getting a similar error:
https://pastebin.com/6QNStUQ1
lelit commentedon Oct 27, 2017
Any chance you could investigate a bit, maybe trying out suggestions above?
thanosd commentedon Oct 27, 2017
Sure thing. The suggestion you are referring to is replacing that include? Giving that a shot right now...
thanosd commentedon Oct 27, 2017
So steps performed:
python setup.py buildSame errors occur. https://pastebin.com/WRt17PaU
Going to dig a bit more on versions of GCC running here, but I'm just jumping into this, so I don't have an intuition of where to look. If someone can throw me some pointers, I can put some time into it.
lelit commentedon Oct 27, 2017
Thank you for trying, I'm thinking about asking advice to the RapidJSON folks...
sontek commentedon Oct 31, 2017
I actually just ran into this on my other Mac... Not sure why but I think it might be related to an older version of clang or something. On my mac that I have no issues with:
on my mac I have this issue:
that is the only difference I've seen so far.
sontek commentedon Oct 31, 2017
I'm using py35, so might be different... The errors are similar:
sontek commentedon Oct 31, 2017
I did:
and still get the errors:
so its not clang.
lelit commentedon Oct 31, 2017
That's so strange! But system/compiler versions may explain why the trouble does not always manifest itself: for example, the Travis build succeeds except with Python 3.4...
sontek commentedon Oct 31, 2017
This is on El Capitan, I'm upgrading to Sierra now, so I'll let you know if that fixes the issue on this system.
Replace Py_IS_NAN() and Py_IS_INFINITY() with std::isnan() and std::i…
lelit commentedon Dec 14, 2017
It seems that #88 fixes the problem, at least on Travis CI: any chance some of you could try it and confirm?
sontek commentedon Dec 15, 2017
Just tested and it works.
lelit commentedon Dec 15, 2017
This should be fixed in current master, I will probably release 0.3 next week or so. Thank you all.
hynek commentedon Dec 15, 2017
Can I interest you into calling it 1.0? 😇
lelit commentedon Dec 20, 2017
I just released v0.3, let's see whether the macOS wheel appears on PyPI.
Wrt v1.0, I'd like to have further attempts at issue #80 before that...
lelit commentedon Dec 20, 2017
Well, it seems there's still something wrong with the MacOS recipe, it fails to find
twine, even if it appears to be installed...Alternative way to execute twine
lelit commentedon Dec 21, 2017
Yay! It seems I ruled it, at last!