Re: Octave on windows testing

[Olaf Till]

On Thu, Mar 10, 2016 at 11:54:21AM -0500, John W. Eaton wrote:
> Why is the struct package overriding the values of CXX and CXXFLAGS?  I
> think it should just use the default values set in mkoctfile.  If a user
> wants to override them, they still can as mkoctfile looks for values of CXX
> and CXXFLAGS in the environment.  But if they are not set in the
> environment, then I don't think the struct package makefile should set them.

Yes, in struct (and optim) this overriding was completely
useless... and anyway before modifying CXXFLAGS mkoctfile should be
queried for them... I'll fix this mistake in these and other packages.

But this also seems to mean that (in package 'database') I can't do:

AC_PROG_CXX
if test "$GXX" != yes; then
AC_MSG_ERROR(...

in configure.ac, but have to check the output of 'mkoctfile -p CXX'
instead. This seems a bit awkward...

On Thu, Mar 10, 2016 at 12:37:50PM -0500, JohnD wrote:
> Currently when cross compiling struct, it is setting CXX to
> i686-w64-mingw32-g++ -std=gnu++11
> 
> Then the CXX=$(CXX) in the Makefile - no quotes - is setting CXX to
> i686-w64-mingw32-g++ with  a stray -std=gnu++11 on the command line
> 
> If I remove both the CXX, CXXFLAGS from preceeding the mkoctfile, it
> compiles ok for me

Thanks for checking, this will be fixed.

Olaf

-- 
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc
Description: Digital signature