Build Question: How to Add -Wl, --option Before Objects In Setup.py?

Cyd Haselton chaselton at gmail.com
Wed Nov 5 16:23:47 CET 2014


On Sun, Nov 2, 2014 at 3:38 PM, Ned Deily <nad at acm.org> wrote:
> In article
> <CAHu5PrYQ3xEgtd-7aHzMDBWk32NPRfxZ24DFdM1oJ-wnmyj8jQ at mail.gmail.com>,
>  Cyd Haselton <chaselton at gmail.com> wrote:
>> Just checking: is sincos() the same as sin() and cos()? Nm output for
>> my toolchain's libm does show sin() and cos() just not sincos()
>
> See, this is what you get when you ask for free help: bad info.

Silly me...what was I thinking?

> sincos
> isn't the same, as a little of googling informs me.  But, as the thread
> starting here indicates:
>

I did a fairly exhaustive search for the sincos() error code I was
getting and that link did not turn up for me at all.
Instead I found a reference that seemed to imply that such a macro
should be implemented but it was not implemented
in GCC-4.8.0.

Then again, the link you found doesn't seem to indicate which version
of GCC has this feature...maybe it is versions
later than 4.8.0...


> https://sourceware.org/ml/binutils/2010-04/msg00219.html
>
> gcc can do an optimization that turns a pair of sin() and cos() calls
> with the same arguments into a single call to sincos().  And there is
> such a pair in Python's complexobject.c.  As the thread explains, there
> are some platforms where sin() and cos() are implemented but not
> sincos().  As suggested in the thread, one approach is to give options
> to gcc to avoid the optimizations:
>
>     CFLAGS='-fno-builtin-sin -fno-builtin-cos'
>
> That does seem to result in a libpython with no references to sincos.

Sounds good...I'll try it if bootstrapping 4.9.0 goes south.
>
> --
>  Ned Deily,
>  nad at acm.org
>
> --
> https://mail.python.org/mailman/listinfo/python-list



More information about the Python-list mailing list