[GiNaC-devel] commit 47a0c68f3e0c4

[Jens Vollinga]

Hi Alexei,

Am 24.05.2011 00:54, schrieb Alexei Sheplyakov:
> tar xaf ginac-1.6.0.tar.bz2
> mkdir build
> cd build
> ../ginac-1.6.0/configure --disable-static
>
>
> This fact makes me think that the check is bogus. Also,

now, at my box there is no warning. I followed the same steps, unpacking 
the tar, doing a vpath build, etc. But no, no warning.

If there were a warning, I would agree that the code in configure.ac 
needs to be fixed.  I would still think it is useful to have a summary 
of possible problems at the end, so my approach would be to fix the 
configure.ac code if necessary, not remove it. But I don't get any bogus 
warning! Does your setup differ significantly from mine (Ubuntu 11.04, x64)?

> ../GiNaC/configure --host=i386-mingw32
>
> on Linux fails due to AC_CHECK_FILE (AC_CHECK_FILE does not work when cross
> compiling, as per autoconf manual). Breaking compilation is not very nice.
> Given that the check itself is not very useful I've decided to remove it.
> As a result it's possible to cross compile GiNaC, and there's no incorrect
> warnings.
>
> Hope this helps,

Yes, it helps.

I overlooked the line in the autoconf manual:

"Be aware that, like most Autoconf macros, they test a feature of the 
host machine, and therefore, they die when cross-compiling."

But it just confuses me: most Autoconf macros fail when cross-compiling? 
REALLY? Why? I mean, why should checking for a file fail if I instruct 
the compiler to use a different assembly language? I don't get it. Since 
I want to keep the extra warning, I am tempted to just replace 
AC_CHECK_FILE with some portable shell code. That should be doable. 
Would you go along that path?

Regards,
Jens