[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] gcc 4.2.2 compile error



MiKRO wrote :
so after all those days it's finally done! I managed to compile
3-stage bootstrap of gnu c and c++ with all libs.

Excellent ! You made it possible !

After that little problem with bash everything went OK except one
thing -- in c_locale.cc (to which points some link in build dir) are
used two functions: strtof(__s, &__sanity) and strtod(__s, &__sanity)
and these functions are unknown at compile time... I have no idea why
this is a problem, cstdlib is included, I just commented out
_GLIBCXX_HAVE_STRTOF / _GLIBCXX_HAVE_STRTOD #defines in this case and
replaced with some fast hack (inspired by #else case of these
conditions).

I remember that I had problems with these.
With the current patch, there is no problem while cross-compiling.
I can't remember exactly, but some define (maybe __BSD ?) was defined during the configure tests, but not while compiling libstdc++-v3, thus causing further problems.

And one cosmetic bug -- I remember I've seen this before (probably by
playing with some sources with ./configure scripts) -- on freemint
host, configure script says our platform is "m68k-atari-mint1.1x"
instead of "m68k-atari-mint". So g++, gcc etc aliases are
"m68k-atari-mint1.16-g++" etc in my case which looks quite strange.
Maybe bug in freemint's information functions?

Go to the GCC source directory and type:
./config.guess
The script tries to guess the host triplet.

The OFFICIAL GCC sources contains:
        echo m68k-atari-mint${UNAME_RELEASE}
with UNAME_RELEASE coming from uname -r

I think the release should not be included (just like for Linux or Cygwin). But if it is here, maybe there is some good reason.
Or just a bad copy/paste ?

--
Vincent Rivière