Re: Cross compler for ppc succeeds on Linux, fails on Cygwin

[Dan Kegel <dank at kegel dot com>]

Don Parsons wrote:
> Using crosstool-0.28-rc7 package, making the cross-compiler was easy on
> Linux for powerpc 7400. Much thanks to Dan and others for this excellent
> package. 
>
> Using the same setup and latest cygwin as of yesterday, it fails early
> in binutil/libiberty when ./configure tests for _doprnt :
>
> ...
> checking for working fork... yes
> checking for working vfork... (cached) yes
> checking for _doprnt...       6 [main] sh 940 cmalloc: cmalloc returned
> NULL
> Signal 11
> make: *** [configure-libiberty] Error 1
>
> Is there an environment variable you could set to say _doprnt answer is
> "no" and skip testing?

You can find this yourself, if you're brave enough, by reading the
configure script.  Let's see... yes, I think

export ac_cv_func__doprnt=no

would do it.  Odd that sh is crashing on you, though.  I haven't seen that.
What operating system are you running?  And are you running cygwin-1.59.1?
Finally, did you download the latest *snapshot* of cygwin1.dll?  It's at
http://cygwin.com/snapshots/cygwin1-20040507.dll.bz2

> Here are two files I created for ppc-7400 (and other dat file I used):
>
> [crosstool-0.28-rc7] #  cat demo-ppc7400.sh
> #!/bin/sh
> set -ex
> TARBALLS_DIR=$HOME/cross_downloads
> RESULT_TOP=/opt/crosstool
> export TARBALLS_DIR RESULT_TOP
> GCC_LANGUAGES="c,c++,f77"
> export GCC_LANGUAGES
>
> mkdir -p /opt/crosstool
> eval `cat powerpc-7400.dat gcc-3.4.0-glibc-2.2.5.dat` sh all.sh --notest
> echo Done.
>
> [crosstool-0.28-rc7] #  cat powerpc-7400.dat
> TARGET=powerpc-7400-linux-gnu
> TARGET_CFLAGS="-O"
> GCC_EXTRA_CONFIG="--with-cpu=7400 --enable-altivec \
> --enable-threads=posix --with-stabs --without-dwarf1 --without-dwarf2"

Thanks.  Can you explain what you needed the option
  --enable-threads=posix
for?  I used to have that, but I think it turned out to cause problems
when compiling with glibc-2.1.3.  And even LFS stopped recommending it:
http://linuxfromscratch.org/pipermail/lfs-dev/2002-September/028385.html

Also, I'm not familiar with stabs format; what made you need to do
  --with-stabs --without-dwarf1 --without-dwarf2
?  (I'm just curious.)

> There seems to be a step overlooked by the crosstools.sh script. 
> Although it does not seem serious.  The binutils package from
> BINUTILS_URL several lines up is intended to be run by doing:
>      rpmbuild -ta binutils-xx.xx.xx.xx.xx.tar.bz2
> and this will apply five patches in its patch dir. via:
>      /bin/sh patches/README
> but crosstool.sh misses this step.  But is usually works anyway or
> maybe the crosstool patches already include these?

Already included; see the last eight lines of getandpatch.sh, starting
with "# binutils-2.14.90.0.3 and up want you to apply a patch".

> Getting back to the _doprnt problem 
>
> In case you can read a stackdump

Nope :-)

> Can anyone suggest a solution or work around for this error?  I could go
> in and take check for _doprnt out but crosstool.sh I think deletes tree
> each time it is run and would remove any changes I made.

Try setting ac_cv_func__doprnt=no and see if that helps.  I worry, though,
that something strange is going on.  I never ran into that in my testing.
I didn't enable Fortran, though, so who knows.

It might be good if you could create a minimal test case for
the doprnt problem, and submit a bug report to Cygwin...
- Dan

--
My technical stuff: http://kegel.com
My politics: see http://www.misleader.org for examples of why I'm for regime change

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com