Cygwin 1.7/1.5 C99 conformance (GCC, libc)

Cygwin 1.7/1.5 C99 conformance (GCC, libc)

[Matthias Andree]

Greetings,

I recently figured that ctrio (from sourceforge.net) fails to link (GCC  
3.4.4 here).

Reason is that ctrio checks the __STDC_VERSION__ for C99 support, which  
there is, and assumes that the implementation were a hosted one (as  
opposed to a freestanding one) and uses powl() without further checks.

Now, Cygwin 1.5 and 1.7 appear to lack long double library support in  
libc/libm -- which is required for a hosted C99 implementation however,  
but not for a freestanding C99 implementation.

Trying to solve this, I found another Cygwin issue: GCC 3.4.4 sets  
__STDC_HOSTED__ to 1 by default (that's a GCC factory default).

I propose that this gets overridden to 0 in Cygwin-ported GCC compilers  
until the library support for a C99-conforming full hosted implementation  
is in place, so that applications can safely check

#if __STDC_HOSTED__ - 0 > 0
/* assume hosted C99 implementation, i. e. with powl() capable libc/libm */
/* ... */
#endif

Thanks.

Best regards

-- 
Matthias Andree

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Cygwin 1.7/1.5 C99 conformance (GCC, libc)

[Dave Korn]

Matthias Andree wrote:

> Now, Cygwin 1.5 and 1.7 appear to lack long double library support in
> libc/libm -- which is required for a hosted C99 implementation however,
> but not for a freestanding C99 implementation.
> 
> Trying to solve this, I found another Cygwin issue: GCC 3.4.4 sets
> __STDC_HOSTED__ to 1 by default (that's a GCC factory default).

  Just FTR, Cygwin-1.5 and gcc-3.4.4 are end-of-life and will never be updated
again.

> I propose that this gets overridden to 0 in Cygwin-ported GCC compilers
> until the library support for a C99-conforming full hosted
> implementation is in place, so that applications can safely check

  Sorry, I don't think that's a good idea.  There are a whole world of other
differences implied by the freestanding-vs-hosted distinction that I think it
would not be wise for the compiler to lie about.  It is indeed unfortunate
that our C library support (which derives from newlib) is incomplete, and this
gap between theory and practice is what autoconf was invented to bridge.
(Autoconf is of course a compromise born of necessity which in an ideal world
wouldn't exist.  But at least it means these problems can be dealt with in a
practical and systematic fashion.)

  It mightn't be too hard to get long double support into newlib, but it's all
a matter of someone having the time and wherewithal to actually do it.

    cheers,
      DaveK

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple