#577 Bootstrap failure with gcc-3.4.0-20040501-1 in __w32_sharedp

[Aaron W. LaFramboise]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

I'm getting a bootstrap failure attempting to compile
MinGW gcc-3.4.0-20040501-1. Specs:
Windows XP with all current patches, Pentium 4 1.4GHz,
128MB RAM
Compiling with MinGW-current, using GCC 3.4RC1:
$ gcc -v
Reading specs from
c:/mingw/bin/../lib/gcc/mingw32/3.4.0/specs
Configured with: ../gcc/configure --with-gcc --with-
gnu-ld --with-gnu-as --host=mingw32 --
target=mingw32 --prefix=/mingw --enable-threads --
disable-nls --enable-
languages=c,c++,f77,ada,objc,java --disable-win32-
registry --disable-shared --enable-sjlj-exceptions --
enable-libgcj --disable-java-awt --without-x --enable-
java-gc=boehm --disable-libgcj-debug --enable-
interpreter --enable-hash-synchronization
Thread model: win32
gcc version 3.4.0 (mingw special)

I get the error when building stage2, here:
./xgcc -B./ -
B/aaronwl/cs/compilers/mingw/gcc/mingw/mingw32/bin/ -
isystem /aaronwl/cs/compilers/mingw/gcc/mingw/mingw3
2/include -
isystem /aaronwl/cs/compilers/mingw/gcc/mingw/mingw3
2/sys-include -
L/aaronwl/cs/compilers/mingw/gcc/build/objdir/gcc/../ld -
O2 -DIN_GCC -W -Wall -Wwrite-strings -Wstrict-
prototypes -Wmissing-prototypes -Wold-style-
definition -isystem ./include -I. -I. -I../../gcc-
3.4.0/gcc -I../../gcc-3.4.0/gcc/. -I../../gcc-
3.4.0/gcc/../include -g0 -finhibit-size-directive -fno-
inline-functions -fno-exceptions -fno-zero-initialized-in-
bss -fno-unit-at-a-time \
-c ../../gcc-3.4.0/gcc/crtstuff.c -DCRT_BEGIN \
-o crtbegin.o

This application has requested the Runtime to terminate
it in an unusual way.
Please contact the application's support team for more
information.
xgcc.exe: Internal error: Aborted (program cc1)
Please submit a full bug report.
See <URL:http://www.mingw.org/bugs.shtml> for
instructions.
make[2]: *** [crtbegin.o] Error 1
make[2]: Leaving directory
`/aaronwl/cs/compilers/mingw/gcc/build/objdir/gcc'
make[1]: *** [stage2_build] Error 2
make[1]: Leaving directory
`/aaronwl/cs/compilers/mingw/gcc/build/objdir/gcc'
make: *** [bootstrap] Error 2

In particular, cc1 is aborting here:
#1 0x006b1b7b in __w32_sharedptr_get (atom=63)
at ../../gcc-3.4.0/gcc/config/i386/w32-shared-
ptr.c:229
228 if (w32_sharedptr->size != sizeof
(W32_EH_SHARED))
229 abort ();

GDB reports:
(gdb) print w32_sharedptr->size
$1 = 56
(gdb) print sizeof(W32_EH_SHARED)
$2 = 60

I'm not familiar with the MinGW sharedptr stuff, so I
can't comment further on the problem at this time.
It may be worth noting that a vanilla FSF gcc-3.4.tar.gz
builds without any problems on the same configuration.

Aaron W. LaFramboise
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFArZZwxDYk2wAn5+8RAhR4AKCSshGaHsDpxhF4LC
RR+FHNx3vMlQCeNB2N
KvxnktVkU2QBqsWnFI0F0O4=
=g4NT
-----END PGP SIGNATURE-----

[Danny Smith]

Logged In: YES
user_id=11494

Thanks for reporting this. I didn't see this, but it is a real bug
that is going to affect many (most?) projects. The ABI has
changed,as Thomas P has already pointed out, with change in
__gthread_mutex_t. I think the w32_atom_suffix string needs
to be amended with addition of an ABI version string (eg "-
v2") to prevent this type of msmatch.

Either that, or we drop support for the shared-ptr stuff until it
gets into official FSF sources.

Danny

[Aaron W. LaFramboise]

Logged In: YES
user_id=1040098

I'm pretty sure this happened because of this:
http://gcc.gnu.org/ml/gcc-patches/2004-06/msg00703.html

Recently, I've noticed that some other people have had
difficult-to-explain GCC bootstrap failures that look like they
might be related to this.

[Aaron W. LaFramboise]

Logged In: YES
user_id=1040098

I'm pretty sure this happened because of this:
http://gcc.gnu.org/ml/gcc-patches/2004-06/msg00703.html

Recently, I've noticed that some other people have had
difficult-to-explain GCC bootstrap failures that look like they
might be related to this.

[Danny Smith]

Logged In: YES
user_id=11494

Actually, if it hadn't been for the MD_STARTFILE_PREFIX
define, the ABI change in w32-shared-ptr.c may not have
been detected so quickly.

I really think that the win32-shared-pr stuff should be either
ditched or made public in FSF forums, but I can't do the latter
because its not my patchset.

Danny

[Aaron W. LaFramboise]

Logged In: YES
user_id=1040098

Now that the startfile patches have been committed, and
have been working correctly for some time, this ticket can be
closed, unless you want to track the w32-shared-ptr ABI
change.