mingw32 target broken

mingw32 target broken

[Adam Megacz]

Hrm, I checked this out a few mintues ago... does anybody know what
the problem might be? I'm building on an i686-gnu-linux host...

/home/megacz/gcc-3.1/bin/gcc/xgcc -B/home/megacz/gcc-3.1/bin/gcc/ -B/usr/mingw32/bin/ -B/usr/mingw32/lib/ -isystem /usr/mingw32/include -O2 -I../../src/gcc/../winsup/include -I../../src/gcc/../winsup/cygwin/include -I../../src/gcc/../winsup/w32api/include -DCROSS_COMPILE -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -isystem ./include   -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -Dinhibit_libc -I. -I. -I../../src/gcc -I../../src/gcc/. -I../../src/gcc/config -I../../src/gcc/../include -DL_chkstk -xassembler-with-cpp -c ../../src/gcc/config/i386/cygwin.asm -o libgcc/./_chkstk.o
/home/megacz/gcc-3.1/bin/gcc/xgcc -B/home/megacz/gcc-3.1/bin/gcc/ -B/usr/mingw32/bin/ -B/usr/mingw32/lib/ -isystem /usr/mingw32/include -O2 -I../../src/gcc/../winsup/include -I../../src/gcc/../winsup/cygwin/include -I../../src/gcc/../winsup/w32api/include -DCROSS_COMPILE -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -isystem ./include   -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -Dinhibit_libc -I. -I. -I../../src/gcc -I../../src/gcc/. -I../../src/gcc/config -I../../src/gcc/../include  -DL_muldi3 -c ../../src/gcc/libgcc2.c -o libgcc/./_muldi3.o
In file included from ../../src/gcc/config/i386/mingw32.h:25,
                 from tconfig.h:4,
                 from ../../src/gcc/libgcc2.c:36:
../../src/gcc/config/i386/cygwin.h:31:19: stdio.h: No such file or directory
In file included from ../../src/gcc/config/i386/mingw32.h:25,
                 from tconfig.h:4,
                 from ../../src/gcc/libgcc2.c:36:
../../src/gcc/config/i386/cygwin.h:435: parse error before '*' token
../../src/gcc/config/i386/cygwin.h:435: warning: function declaration isn't a prototype
../../src/gcc/config/i386/cygwin.h:437: parse error before '*' token
../../src/gcc/config/i386/cygwin.h:437: warning: function declaration isn't a prototype
make[2]: *** [libgcc/./_muldi3.o] Error 1
make[2]: Leaving directory `/home/megacz/gcc-3.1/bin/gcc'
make[1]: *** [libgcc.a] Error 2
make[1]: Leaving directory `/home/megacz/gcc-3.1/bin/gcc'
make: *** [all-gcc] Error 2

Re: mingw32 target broken

[Adam Megacz]

Adam Megacz <gcc@lists.megacz.com> writes:
> Hrm, I checked this out a few mintues ago... does anybody know what
> the problem might be? I'm building on an i686-gnu-linux host...

My configure line was

  ../src/configure --enable-threads=posix --prefix=/usr --enable-shared --enable-languages=c++,java,c --disable-nls --target=mingw

if that helps...

  - a

Re: mingw32 target broken [cygwin as well]

[Adam Megacz]

Craig Rodrigues <rodrigc@mediaone.net> writes:
> Post the output of the configure, and the config.log that was created as well.

No config.log was created. I have included config.status below...

Side note: this seems to happen with target=i386-pc-mingw and
target=i386-pc-cygwin as well. For those just joining the thread, I'm
using a fresh cvs checkout from a few hours ago.

== configure command ===========================================================

../src/configure --enable-threads=posix --prefix=/usr --enable-shared \
                 --enable-languages=c++,java,c --disable-nls --target=mingw

== configure output ============================================================

Configuring for a i686-pc-linux-gnu host.
*** This configuration is not supported in the following subdirectories:
     target-libffi target-boehm-gc target-zlib target-libjava target-libchill target-libstdc++-v3 target-libf2c zlib fastjar target-libobjc
    (Any other directories should still work fine.)
Created "Makefile" in /home/megacz/gcc-3.1/bin using "mh-frag" and "mt-frag"
Configuring libiberty...
creating cache ../config.cache
checking host system type... i686-pc-linux-gnu
checking build system type... i686-pc-linux-gnu
checking for ar... ar
checking for ranlib... ranlib
checking for gcc... gcc
checking whether we are using GNU C... yes
checking whether gcc accepts -g... yes
checking for POSIXized ISC... no
checking for working const... yes
checking for inline... inline
checking for a BSD compatible install... /usr/bin/install -c
Appending ../../src/libiberty/config/../../config/mh-x86pic to xhost-mkfrag
checking how to run the C preprocessor... gcc -E
checking for sys/file.h... yes
checking for sys/param.h... yes
checking for limits.h... yes
checking for stdlib.h... yes
checking for string.h... yes
checking for unistd.h... yes
checking for strings.h... yes
checking for sys/time.h... yes
checking for time.h... yes
checking for sys/resource.h... yes
checking for sys/stat.h... yes
checking for sys/mman.h... yes
checking for fcntl.h... yes
checking for sys/wait.h that is POSIX.1 compatible... yes
checking whether time.h and sys/time.h may both be included... yes
checking whether errno must be declared... no
checking whether the C compiler (gcc -g -O2 ) works... yes
checking whether the C compiler (gcc -g -O2 ) is a cross-compiler... no
checking for asprintf... yes
checking for atexit... yes
checking for basename... yes
checking for bcmp... yes
checking for bcopy... yes
checking for bsearch... yes
checking for bzero... yes
checking for calloc... yes
checking for clock... yes
checking for ffs... yes
checking for getcwd... yes
checking for getpagesize... yes
checking for index... yes
checking for insque... yes
checking for memchr... yes
checking for memcmp... yes
checking for memcpy... yes
checking for memmove... yes
checking for memset... yes
checking for mkstemps... no
checking for putenv... yes
checking for random... yes
checking for rename... yes
checking for rindex... yes
checking for setenv... yes
checking for sigsetmask... yes
checking for strcasecmp... yes
checking for strchr... yes
checking for strdup... yes
checking for strncasecmp... yes
checking for strrchr... yes
checking for strstr... yes
checking for strtod... yes
checking for strtol... yes
checking for strtoul... yes
checking for tmpnam... yes
checking for vasprintf... yes
checking for vfprintf... yes
checking for vprintf... yes
checking for vsprintf... yes
checking for waitpid... yes
checking whether alloca needs Cray hooks... no
checking stack direction for C alloca... -1
checking for ANSI C header files... yes
checking for pid_t... yes
checking for vfork.h... no
checking for working vfork... yes
checking for sys_errlist... yes
checking for sys_nerr... yes
checking for sys_siglist... yes
checking for getrusage... yes
checking for on_exit... yes
checking for psignal... yes
checking for strerror... yes
checking for strsignal... yes
checking for sysconf... yes
checking for times... yes
checking for sbrk... yes
checking for gettimeofday... yes
checking for unistd.h... (cached) yes
checking for getpagesize... (cached) yes
checking for working mmap... yes
checking for working strncmp... yes
updating cache ../config.cache
creating ./config.status
creating Makefile
creating testsuite/Makefile
creating config.h
Configuring gcc...
loading cache ../config.cache
checking LIBRARY_PATH variable... ok
checking GCC_EXEC_PREFIX variable... ok
checking host system type... i686-pc-linux-gnu
checking target system type... i386-pc-mingw32
checking build system type... i686-pc-linux-gnu
checking for gcc... (cached) gcc
checking whether the C compiler (gcc -g -O2 ) works... yes
checking whether the C compiler (gcc -g -O2 ) is a cross-compiler... no
checking whether we are using GNU C... (cached) yes
checking whether gcc accepts -g... (cached) yes
checking whether gcc and cc understand -c and -o together... yes
checking whether gcc accepts -Wno-long-long... yes
checking how to run the C preprocessor... (cached) gcc -E
checking for inline... (cached) inline
checking for volatile... yes
checking for long double... yes
checking for long long int... yes
checking for __int64... no
checking for built-in _Bool... yes
checking size of short... 2
checking size of int... 4
checking size of long... 4
checking size of long long... 8
checking execution character set... ASCII
checking whether make sets ${MAKE}... yes
checking whether a default assembler was specified... no
checking whether a default linker was specified... no
checking for GNU C library... yes
checking for gawk... no
checking for mawk... mawk
checking whether ln works... yes
checking whether ln -s works... yes
checking for ranlib... (cached) ranlib
checking for a BSD compatible install... (cached) /usr/bin/install -c
checking for ANSI C header files... (cached) yes
checking whether time.h and sys/time.h may both be included... (cached) yes
checking for working stdbool.h... yes
checking whether string.h and strings.h may both be included... yes
checking for sys/wait.h that is POSIX.1 compatible... (cached) yes
checking for limits.h... (cached) yes
checking for stddef.h... yes
checking for string.h... (cached) yes
checking for strings.h... (cached) yes
checking for stdlib.h... (cached) yes
checking for time.h... (cached) yes
checking for fcntl.h... (cached) yes
checking for unistd.h... (cached) yes
checking for sys/file.h... (cached) yes
checking for sys/time.h... (cached) yes
checking for sys/resource.h... (cached) yes
checking for sys/param.h... (cached) yes
checking for sys/times.h... yes
checking for sys/stat.h... (cached) yes
checking for direct.h... no
checking for malloc.h... yes
checking for langinfo.h... yes
checking for thread.h... no
checking for pthread.h... yes
checking for CHAR_BIT... yes
checking byte ordering... little-endian
checking floating point format... IEEE (little-endian)
checking for gnatbind... no
checking for mktemp... yes
checking for makeinfo... makeinfo
checking for modern makeinfo... yes
checking for recent Pod::Man... yes
checking for flex... flex
checking for bison... bison
checking for collect2 libraries... none required
checking for library containing exc_resume... no
checking for preprocessor stringizing operator... yes
checking for inttypes.h... yes
checking for strtoul... (cached) yes
checking for bsearch... (cached) yes
checking for popen... yes
checking for times... (cached) yes
checking for clock... (cached) yes
checking for strchr... (cached) yes
checking for strrchr... (cached) yes
checking for kill... yes
checking for getrlimit... yes
checking for setrlimit... yes
checking for atoll... yes
checking for atoq... no
checking for sysconf... (cached) yes
checking for isascii... yes
checking for gettimeofday... (cached) yes
checking for strsignal... (cached) yes
checking for putc_unlocked... yes
checking for fputc_unlocked... yes
checking for fputs_unlocked... yes
checking for getrusage... (cached) yes
checking for nl_langinfo... yes
checking for lstat... yes
checking for ssize_t... yes
checking for uid_t in sys/types.h... yes
checking type of array argument to getgroups... gid_t
checking for vprintf... (cached) yes
checking for strstr... (cached) yes
checking whether the printf functions support %p... yes
checking for pid_t... (cached) yes
checking for vfork.h... (cached) no
checking for working vfork... (cached) yes
checking for getpagesize... (cached) yes
checking for working mmap from /dev/zero... yes
checking for working mmap with MAP_ANON(YMOUS)... yes
checking for working mmap of a file... yes
checking for iconv... yes
checking for iconv declaration... 
         extern size_t iconv (iconv_t cd, char * *inbuf, size_t *inbytesleft, char * *outbuf, size_t *outbytesleft);
checking whether getenv is declared... yes
checking whether atol is declared... yes
checking whether sbrk is declared... yes
checking whether abort is declared... yes
checking whether atof is declared... yes
checking whether getcwd is declared... yes
checking whether getwd is declared... yes
checking whether strsignal is declared... yes
checking whether putc_unlocked is declared... yes
checking whether fputs_unlocked is declared... yes
checking whether strstr is declared... yes
checking whether environ is declared... yes
checking whether errno is declared... yes
checking whether malloc is declared... yes
checking whether realloc is declared... yes
checking whether calloc is declared... yes
checking whether free is declared... yes
checking whether basename is declared... yes
checking whether getopt is declared... yes
checking whether clock is declared... yes
checking whether getrlimit is declared... yes
checking whether setrlimit is declared... yes
checking whether getrusage is declared... yes
checking whether times is declared... yes
checking for struct tms... yes
checking for clock_t... yes
checking if mkdir takes one argument... no
Using `../../src/gcc/config/i386/i386.c' for machine-specific logic.
Using `../../src/gcc/config/i386/i386.md' as machine description file.
Using the following target machine macro files:
	../../src/gcc/config/i386/mingw32.h
	../../src/gcc/config/i386/crtdll.h
checking for strerror in -lcposix... no
checking for working const... (cached) yes
checking for off_t... yes
checking for size_t... yes
checking for argz.h... yes
checking for limits.h... (cached) yes
checking for locale.h... yes
checking for nl_types.h... yes
checking for malloc.h... (cached) yes
checking for string.h... (cached) yes
checking for unistd.h... (cached) yes
checking for sys/param.h... (cached) yes
checking for getcwd... (cached) yes
checking for munmap... yes
checking for putenv... (cached) yes
checking for setenv... (cached) yes
checking for setlocale... yes
checking for strchr... (cached) yes
checking for strcasecmp... (cached) yes
checking for strdup... (cached) yes
checking for __argz_count... yes
checking for __argz_stringify... yes
checking for __argz_next... yes
checking for stpcpy... yes
checking for LC_MESSAGES... yes
checking whether NLS is requested... no
checking what assembler to use... 
checking what nm to use... 
checking assembler alignment features... none
checking assembler subsection support... no
checking assembler weak support... no
checking assembler hidden support... no
checking assembler leb128 support... no
checking assembler eh_frame optimization... no
checking assembler instructions... 
checking assembler dwarf2 debug_line support... no
Using ggc-page for garbage collection.
checking whether to enable maintainer-specific portions of Makefiles... no
Links are now set up to build a cross-compiler for i386-pc-mingw32
  from i686-pc-linux-gnu.
updating cache ../config.cache
creating ./config.status
creating Makefile
creating intl/Makefile
creating po/Makefile.in
creating fixinc/Makefile
creating gccbug
creating mklibgcc
creating auto-host.h
linking ../../src/gcc/intl/libgettext.h to intl/libintl.h
creating libintl.h

== config.status ===============================================================

#!/bin/sh
# This file was generated automatically by configure.  Do not edit.
# This directory was configured as follows:
../src/configure --with-gcc-version-trigger=/home/megacz/gcc-3.1/src/gcc/version.c --host=i686-pc-linux-gnu --enable-threads=posix --prefix=/usr --enable-shared --enable-languages=c --disable-nls --target=mingw32 --norecursion 
# using "mh-frag" and "mt-frag"



== build errors ================================================================
...
/home/megacz/gcc-3.1/bin/gcc/xgcc -B/home/megacz/gcc-3.1/bin/gcc/ -B/usr/mingw32/bin/ -B/usr/mingw32/lib/ -isystem /usr/mingw32/include -O2 -I../../src/gcc/..\/winsup/include -I../../src/gcc/../winsup/cygwin/include -I../../src/gcc/../winsup/w32api/include -DCROSS_COMPILE -DIN_GCC    -W -Wall -Wwrite-strings -Wstric\t-prototypes -Wmissing-prototypes -isystem ./include   -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -Dinhibit_libc -I. -I. -I../../src/gcc -I\../../src/gcc/. -I../../src/gcc/config -I../../src/gcc/../include -DL_chkstk -xassembler-with-cpp -c ../../src/gcc/config/i386/cygwin.asm -o libgcc/./_chkstk.\o
/home/megacz/gcc-3.1/bin/gcc/xgcc -B/home/megacz/gcc-3.1/bin/gcc/ -B/usr/mingw32/bin/ -B/usr/mingw32/lib/ -isystem /usr/mingw32/include -O2 -I../../src/gcc/..\/winsup/include -I../../src/gcc/../winsup/cygwin/include -I../../src/gcc/../winsup/w32api/include -DCROSS_COMPILE -DIN_GCC    -W -Wall -Wwrite-strings -Wstric\t-prototypes -Wmissing-prototypes -isystem ./include   -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -Dinhibit_libc -I. -I. -I../../src/gcc -I\../../src/gcc/. -I../../src/gcc/config -I../../src/gcc/../include  -DL_muldi3 -c ../../src/gcc/libgcc2.c -o libgcc/./_muldi3.o
In file included from ../../src/gcc/config/i386/mingw32.h:25,
                 from tconfig.h:4,
                 from ../../src/gcc/libgcc2.c:36:
../../src/gcc/config/i386/cygwin.h:31:19: stdio.h: No such file or directory
In file included from ../../src/gcc/config/i386/mingw32.h:25,
                 from tconfig.h:4,
                 from ../../src/gcc/libgcc2.c:36:
../../src/gcc/config/i386/cygwin.h:435: parse error before '*' token
../../src/gcc/config/i386/cygwin.h:435: warning: function declaration isn't a prototype
../../src/gcc/config/i386/cygwin.h:437: parse error before '*' token
../../src/gcc/config/i386/cygwin.h:437: warning: function declaration isn't a prototype
make[2]: *** [libgcc/./_muldi3.o] Error 1
make[2]: Leaving directory `/home/megacz/gcc-3.1/bin/gcc'
make[1]: *** [libgcc.a] Error 2
make[1]: Leaving directory `/home/megacz/gcc-3.1/bin/gcc'
make: *** [all-gcc] Error 2

Re: mingw32 target broken [cygwin as well]

[Craig Rodrigues]

Hi,

I don't think building a mingw cross compiler is a
straightforward thing.  You need the header files
for the mingw runtime and for the w32api.

See:
http://members.telering.at/jessich/mingw/index.html
http://sources.redhat.com/ml/crossgcc/2001-q4/msg00310.html

And you might want to ask on the mingw mailing lists
at http://www.mingw.org
-- 
Craig Rodrigues        
http://www.gis.net/~craigr    
rodrigc@mediaone.net          

Re: mingw32 target broken [cygwin as well] [didn't know cross-compilers were such an ordeal]

[Adam Megacz]

Craig Rodrigues <rodrigc@mediaone.net> writes:
> http://members.telering.at/jessich/mingw/index.html
> http://sources.redhat.com/ml/crossgcc/2001-q4/msg00310.html

Wow, I just checked out that patching script -- it's enormous!

Why is it so hard to build a cross compiler? Shouldn't it be just like
a regular compiler build, except that instead of using system
libraries (/usr/lib), it uses alternate libraries
(/usr/i686-pc-mingw)? AFAIK, what platform a compiler binary is
compiled for really shouldn't have anything to do with what kind of
binaries it puts out... it's not like the compiler copies bits of
itself into the output...

Oh well, I'm going to quit whining right now and ssh to a win98
machine running cygwin to do my builds.

Thanks for the help...

  - a

Re: mingw32 target broken [cygwin as well]

[Jeff Sturm]

On 8 Dec 2001, Adam Megacz wrote:
> Side note: this seems to happen with target=i386-pc-mingw and
> target=i386-pc-cygwin as well. For those just joining the thread, I'm
> using a fresh cvs checkout from a few hours ago.

I haven't had much trouble building a cross compiler for mingw/cygwin in
the past.  Note however that gcc needs to find the target headers in
$prefix/<target>/include, so you'll want to copy these before configuring
gcc.  (The CrossGCC faq hints that newlib targets can be built in one
pass however, so cygwin might be a little easier starting from a unified
tree.)

> -isystem /usr/mingw32/include

This is where gcc is looking for stdio.h...

> ../../src/gcc/config/i386/cygwin.h:31:19: stdio.h: No such file or directory

...which is included from here.  Looks like /usr/mingw32/include/stdio.h
doesn't exist on your build machine.

Jeff

Re: mingw32 target broken [cygwin as well] [didn't know cross-compilers were such an ordeal]

[Jeff Sturm]

On 8 Dec 2001, Adam Megacz wrote:
> 
> Craig Rodrigues <rodrigc@mediaone.net> writes:
> > http://members.telering.at/jessich/mingw/index.html
> > http://sources.redhat.com/ml/crossgcc/2001-q4/msg00310.html
> 
> Wow, I just checked out that patching script -- it's enormous!

Last time I looked at the mingw patches, it was mostly features that are
not essential to building gcc, but desirable for windows users, e.g.
anonymous structs.  I've successfully built C cross compilers without any
patches.  (C++ may be problematic, especially for DLLs.)

> Oh well, I'm going to quit whining right now and ssh to a win98
> machine running cygwin to do my builds.

I may as well warn you that if you are doing any gcj work, there is a bug
that prevents libgcj from bootstrapping on machines without case-sensitive
filesystems (OSX on HFS, all Windows releases).

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=2388

(If you find out that this is fixed on the trunk, let us know so we can
close the PR!)

A Linux-hosted cross compiler doesn't have this problem.  (Besides, do you
_really_ want to watch libtool on cygwin?)

Jeff

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Adam Megacz]

Jeff Sturm <jsturm@one-point.com> writes:
> Note however that gcc needs to find the target headers in
> $prefix/<target>/include, so you'll want to copy these before
> configuring gcc

Ah, this was the problem. I also had to compile a set of
cross-binutils. This time I got much farther -- it seems to get most
of the C compiler to compile, but I'm running into problems with
libstdc++... has anybody seen anything like this before?

../src/configure \
                 --prefix=/usr \
                 --enable-shared \
                 --enable-languages=c,c++,java \
                 --disable-nls \
                 --with-as=/usr/cross-binutils/i686-pc-mingw32/bin/as \
                 --target=i686-pc-mingw32 \
                 --with-gnu-ld \
                 --with-gnu-as \
                 --with-ld=/usr/cross-binutils/i686-pc-mingw32/bin/ld

/home/megacz/gcc-3.1/bin/gcc/xgcc -B/home/megacz/gcc-3.1/bin/gcc/ -nostdinc++ -L/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/src -L/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/src/.libs -B/usr/i686-pc-mingw32/bin/ -B/usr/i686-pc-mingw32/lib/ -isystem /usr/i686-pc-mingw32/include -nostdinc++ -I/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32 -I/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include -I../../../../src/libstdc++-v3/libsupc++ -I../../../../src/libstdc++-v3/libmath -g -O2 -fno-implicit-templates -Wall -Wno-format -W -Wwrite-strings -Winline -fdiagnostics-show-location=once -g -c basic_file.cc -o basic_file.o
In file included from /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/bits/locale_facets.h:54,
                 from /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/bits/basic_ios.h:41,
                 from /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/bits/std_ios.h:51,
                 from /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/bits/basic_file.h:45,
                 from basic_file.cc:34:
/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:46: `
   _U' was not declared in this scope
/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:47: `
   _L' was not declared in this scope
/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:48: `
   _U' was not declared in this scope
/home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:48: `
   _L' was not declared in this scope

Re: mingw32 target broken [cygwin as well] [didn't know cross-compilers were such an ordeal]

[Adam Megacz]

Jeff Sturm <jsturm@one-point.com> writes:
> > Craig Rodrigues <rodrigc@mediaone.net> writes:
> > > http://members.telering.at/jessich/mingw/index.html
> > > http://sources.redhat.com/ml/crossgcc/2001-q4/msg00310.html

> I may as well warn you that if you are doing any gcj work, there is
> a bug that prevents libgcj from bootstrapping on machines without
> case-sensitive filesystems (OSX on HFS, all Windows releases).

Ugh, that's really bad... I was hoping to try for a cross compiler,
and if that fails, fall back to ssh'ing to a cygwin machine for
builds. Now it looks like the cross-compiler is my only hope...

I've got from now until 16-Dec to get libgcj running on mingw. When I
had hoped that this would be a simple-though-tedious process of
pairing up java.* methods to Win32 API calls, with appropriate glue
code... so far I've spent two days just trying to get a working
compiler =)

  - a

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Craig Rodrigues]

On Sat, Dec 08, 2001 at 07:01:06PM -0500, Adam Megacz wrote:
> 
>    _U' was not declared in this scope
> /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:47: `
>    _L' was not declared in this scope
> /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:48: `
>    _U' was not declared in this scope
> /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:48: `
>    _L' was not declared in this scope

Hi,

Check out this thread on the libstdc++ mailing list:

http://gcc.gnu.org/ml/libstdc++/2001-03/msg00131.html
http://gcc.gnu.org/ml/libstdc++/2001-03/msg00134.html
http://gcc.gnu.org/ml/libstdc++/2001-03/msg00143.html

and this patch from bkoz:
http://gcc.gnu.org/ml/libstdc++/2001-03/msg00144.html

-- 
Craig Rodrigues        
http://www.gis.net/~craigr    
rodrigc@mediaone.net          

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Adam Megacz]

Craig Rodrigues <rodrigc@mediaone.net> writes:
> >    _U' was not declared in this scope
> > /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:47: `
> >    _L' was not declared in this scope
> > /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:48: `
> >    _U' was not declared in this scope
> > /home/megacz/gcc-3.1/bin/i686-pc-mingw32/libstdc++-v3/include/i686-pc-mingw32/bits/ctype_base.h:48: `
> >    _L' was not declared in this scope
> 
> and this patch from bkoz:
> http://gcc.gnu.org/ml/libstdc++/2001-03/msg00144.html

A bit stale, but I was able to hand-edit for the same effect, and it
worked.

I'm not too familiar with the administrative end of the gcc project --
is there some patch-approval process that patches like this have not
gone through? Is that why they don't end up getting back into the
trunk? Otherwise, may I post a diff to the latest CVS in the hopes of
getting it folded in?

I've also noticed that the configure in the root directory of the gcc
checkout does not pass the --target=$TARGET option to the invocation
of configure for libstdc++-v3, although it does pass --with-target-dir
-- can anybody verify this and/or advise me on how to fix it "the
right way"? (right now I'm just manually re-running configure for
libstdc++-v3).

  - a

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Craig Rodrigues]

On Sat, Dec 08, 2001 at 09:20:37PM -0500, Adam Megacz wrote:
> I'm not too familiar with the administrative end of the gcc project --
> is there some patch-approval process that patches like this have not
> gone through? Is that why they don't end up getting back into the
> trunk? Otherwise, may I post a diff to the latest CVS in the hopes of
> getting it folded in?

The gcc patch submission guidelines are here:
http://gcc.gnu.org/contribute.html

If you have a patch that solves a problem, then definitely
contribute it.

The GCC patch review process is not always automatic though, 
because the people involved with GCC development are very busy,
so you sometimes need to send out reminder e-mails.
-- 
Craig Rodrigues        
http://www.gis.net/~craigr    
rodrigc@mediaone.net          

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Adam Megacz]

Craig Rodrigues <rodrigc@mediaone.net> writes:
> If you have a patch that solves a problem, then definitely
> contribute it.

Sounds good. If/when I finally get a working cross compiler, I'll
build up a patch and try to get it accepted.

I'm going to be releasing an open source project on 01-Jan-02 that
targets mingw, so it'd be really nice if I could tell people
interested in building it to just "cvs -z9 co -D
$SOME_DATE_KNOWN_TO_WORK gcc" instead of having to go through a long
list of patches/hacks...

  - a

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Jeff Sturm]

On 8 Dec 2001, Adam Megacz wrote:
> > http://gcc.gnu.org/ml/libstdc++/2001-03/msg00144.html
> 
> A bit stale, but I was able to hand-edit for the same effect, and it
> worked.

A dirty little secret: if all you want is the java runtime, you can skip
building libstdc++.  Just remove it from the toplevel Makefile.

Unfortunately, libjava assumes cross compilers target newlib, just as
libstdc++ does...

if test -n "${with_cross_host}"; then
   # We are being configured with a cross compiler.  AC_REPLACE_FUNCS
   # may not work correctly, because the compiler may not be able to
   # link executables.

   # We assume newlib.  This lets us hard-code the functions we know
   # we'll have.

This never seemed like desirable behavior to me, in part because newlib
isn't even GNU software.  I'd prefer that configure attempt to link a
program before it assumes it cannot.

> I've also noticed that the configure in the root directory of the gcc
> checkout does not pass the --target=$TARGET option to the invocation
> of configure for libstdc++-v3, although it does pass --with-target-dir

In the target subdirs, "target" becomes "host".

Jeff

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Adam Megacz]

Jeff Sturm <jsturm@one-point.com> writes:
> A dirty little secret: if all you want is the java runtime, you can skip
> building libstdc++.  Just remove it from the toplevel Makefile.
> 
> Unfortunately, libjava assumes cross compilers target newlib, just as
> libstdc++ does...
> 
> if test -n "${with_cross_host}"; then
>    # We are being configured with a cross compiler.  AC_REPLACE_FUNCS
>    # may not work correctly, because the compiler may not be able to
>    # link executables.
> 
>    # We assume newlib.  This lets us hard-code the functions we know
>    # we'll have.


Okay, I think I've worked around this for libstdc++-v3... can somebody
with more gcc-hacking experience let me know if this is too kludgey to
submit to gcc-patches@gcc.gnu.org?

I basically just special-cased out mingw32 with the stuff that I know
it doesn't have (said knowledge was acquired by the highly scientific
process of trial and error until it built). I'd like to do a more
elegant job, but I don't really understand gcc's autoconf setup to
come up with anything cleaner.

I've also put this patch into http://www.xwt.org/creating.a.mingw.crosscompiler.html

  - a


Index: configure.in
===================================================================
RCS file: /cvs/gcc/gcc/libstdc++-v3/configure.in,v
retrieving revision 1.79
diff -c -3 -p -r1.79 configure.in
*** configure.in	2001/12/03 22:28:57	1.79
--- configure.in	2001/12/10 04:28:53
*************** if test -n "$with_cross_host" || test x"
*** 124,129 ****
--- 124,132 ----
        AC_DEFINE(HAVE_SINCOS)
        AC_DEFINE(HAVE_SINCOSF)
        ;;
+     *-mingw*)
+       os_include_dir="config/os/mingw32"
+       ;;
      *)
        os_include_dir="config/os/newlib"
        AC_DEFINE(HAVE_HYPOT)
*************** if test -n "$with_cross_host" || test x"
*** 136,171 ****
    # AC_FUNC_MMAP
    AC_DEFINE(HAVE_MMAP)
  
!   AC_DEFINE(HAVE_ACOSF)
!   AC_DEFINE(HAVE_ASINF)
!   AC_DEFINE(HAVE_ATAN2F)
!   AC_DEFINE(HAVE_ATANF)
!   AC_DEFINE(HAVE_CEILF)
!   AC_DEFINE(HAVE_COPYSIGN)
!   AC_DEFINE(HAVE_COPYSIGNF)
!   AC_DEFINE(HAVE_COSF)
!   AC_DEFINE(HAVE_COSHF)
!   AC_DEFINE(HAVE_EXPF)
!   AC_DEFINE(HAVE_FABSF)
!   AC_DEFINE(HAVE_FINITE)
!   AC_DEFINE(HAVE_FINITEF)
!   AC_DEFINE(HAVE_FLOORF)
!   AC_DEFINE(HAVE_FMODF)
!   AC_DEFINE(HAVE_FREXPF)
!   AC_DEFINE(HAVE_ISINF)
!   AC_DEFINE(HAVE_ISINFF)
!   AC_DEFINE(HAVE_ISNAN)
!   AC_DEFINE(HAVE_ISNANF)
!   AC_DEFINE(HAVE_LDEXPF)
!   AC_DEFINE(HAVE_LOG10F)
!   AC_DEFINE(HAVE_LOGF)
!   AC_DEFINE(HAVE_MODFF)
!   AC_DEFINE(HAVE_POWF)
!   AC_DEFINE(HAVE_SINF)
!   AC_DEFINE(HAVE_SINHF)
!   AC_DEFINE(HAVE_SQRTF)
!   AC_DEFINE(HAVE_TANF)
!   AC_DEFINE(HAVE_TANHF)
  
    # At some point, we should differentiate between architectures
    # like x86, which have long double versions, and alpha/powerpc/etc.,
--- 139,179 ----
    # AC_FUNC_MMAP
    AC_DEFINE(HAVE_MMAP)
  
!   case "$target_alias" in
!     *-mingw*)
!       ;;
!     *)
!       AC_DEFINE(HAVE_ACOSF)
!       AC_DEFINE(HAVE_ASINF)
!       AC_DEFINE(HAVE_ATAN2F)
!       AC_DEFINE(HAVE_ATANF)
!       AC_DEFINE(HAVE_CEILF)
!       AC_DEFINE(HAVE_COPYSIGN)
!       AC_DEFINE(HAVE_COPYSIGNF)
!       AC_DEFINE(HAVE_COSF)
!       AC_DEFINE(HAVE_COSHF)
!       AC_DEFINE(HAVE_EXPF)
!       AC_DEFINE(HAVE_FABSF)
!       AC_DEFINE(HAVE_FINITE)
!       AC_DEFINE(HAVE_FINITEF)
!       AC_DEFINE(HAVE_FLOORF)
!       AC_DEFINE(HAVE_FMODF)
!       AC_DEFINE(HAVE_FREXPF)
!       AC_DEFINE(HAVE_ISINF)
!       AC_DEFINE(HAVE_ISINFF)
!       AC_DEFINE(HAVE_ISNAN)
!       AC_DEFINE(HAVE_ISNANF)
!       AC_DEFINE(HAVE_LDEXPF)
!       AC_DEFINE(HAVE_LOG10F)
!       AC_DEFINE(HAVE_LOGF)
!       AC_DEFINE(HAVE_MODFF)
!       AC_DEFINE(HAVE_POWF)
!       AC_DEFINE(HAVE_SINF)
!       AC_DEFINE(HAVE_SINHF)
!       AC_DEFINE(HAVE_SQRTF)
!       AC_DEFINE(HAVE_TANF)
!       AC_DEFINE(HAVE_TANHF)
!   esac
  
    # At some point, we should differentiate between architectures
    # like x86, which have long double versions, and alpha/powerpc/etc.,

Re: mingw32 target broken [cygwin as well] [the saga continues]

[DJ Delorie]

> This never seemed like desirable behavior to me, in part because newlib
> isn't even GNU software.  I'd prefer that configure attempt to link a
> program before it assumes it cannot.

Newlib is an operating system, just like solaris or IRIX.  It doesn't
have to be GNU software for us to support it, it just has to support
and allow GNU software.  And you can't always do a test link with
cross compilers, because you may not have built enough support stuff
(crt0, libc, gas/ld) to do so.  In fact, for a canadian cross, you
can't link *at all* because you just can't run the linker.

Newlib is a popular target for GNU tools, so people have contributed
extra support for it.  There are other platforms we do this for, like
cygwin and vxworks.

> > I've also noticed that the configure in the root directory of the gcc
> > checkout does not pass the --target=$TARGET option to the invocation
> > of configure for libstdc++-v3, although it does pass --with-target-dir
> 
> In the target subdirs, "target" becomes "host".

Yeah, that confuses people, but it does make sense.  "host" is what
you're building *for*.  For target libraries, you're building for the
--target, so $host is --target.

$build is what you're building *on*.
$host is what you're building *for*.
$target is what the stuff you're building produces code for.

With these definitions, it's meaningless to use $target with a
library, because libraries do not themselves produce code.

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Joseph S. Myers]

On Mon, 10 Dec 2001, DJ Delorie wrote:

> (crt0, libc, gas/ld) to do so.  In fact, for a canadian cross, you
> can't link *at all* because you just can't run the linker.

But for a Canadian cross, surely you should have a full build x target
toolchain - including the linker and target libraries - installed (as well
as a build x host one) before building the Canadian cross?  (At least,
etc/configure.texi in src says so.)

-- 
Joseph S. Myers
jsm28@cam.ac.uk

Re: mingw32 target broken [cygwin as well] [the saga continues]

[DJ Delorie]

> But for a Canadian cross, surely you should have a full build x target
> toolchain - including the linker and target libraries - installed (as well
> as a build x host one) before building the Canadian cross?  (At least,
> etc/configure.texi in src says so.)

There's no guarantee it's the same version as the library you're
building though.  But you're right, we should have a working linker.
It's *executing* the tests that you can't rely on.

And for Cygwin it's even worse, as parts of libiberty.a are used to
build cygwin1.dll, so if you did the linking test you wouldn't build
the very functions cygwin pulls from libiberty to build the dll!

Re: mingw32 target broken [cygwin as well] [the saga continues]

[Jeff Sturm]

On Mon, 10 Dec 2001, DJ Delorie wrote:
> Newlib is an operating system, just like solaris or IRIX.  It doesn't
> have to be GNU software for us to support it, it just has to support
> and allow GNU software.

Yeah.  It's just that assuming a non-GNU runtime doesn't seem to fit the
ideology.  (Not that GNU has anything better for embedded use... at least
newlib is free software.)

> And you can't always do a test link with
> cross compilers, because you may not have built enough support stuff
> (crt0, libc, gas/ld) to do so.

But it helps if you do.  When I build a cross compiler, I typically follow
these steps:

1) Install target runtime headers.
2) Build/install cross binutils.
3) Build/install C cross compiler.
4) Build/install target runtime libraries.
5) Build/install other languages (c++, java).

At step 5) I have everything I need for AC_TRY_LINK.  I could
save a lot of time if configure would first attempt to link, falling back
on the newlib guesses.

> > In the target subdirs, "target" becomes "host".
> 
> Yeah, that confuses people, but it does make sense.  "host" is what
> you're building *for*.  For target libraries, you're building for the
> --target, so $host is --target.

That's how I understood it.  But I looked in libstdc++-v3 and libjava...
the former has a configure.target script, the latter configure.host.  So
that interpretation isn't uniform throughout GCC.

Jeff

Re: mingw32 target broken [cygwin as well] [the saga continues]

[DJ Delorie]

> > Yeah, that confuses people, but it does make sense.  "host" is what
> > you're building *for*.  For target libraries, you're building for the
> > --target, so $host is --target.
> 
> That's how I understood it.  But I looked in libstdc++-v3 and libjava...
> the former has a configure.target script, the latter configure.host.  So
> that interpretation isn't uniform throughout GCC.

I *told* you it confuses people ;-)