https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88772
--- Comment #4 from Andoni <ylatuya at gmail dot com> ---
(In reply to Eric Botcazou from comment #3)
> > I just wiped the build to start a clean build from scratch, but I remember
> > checking this and it was "no". I can confirm it in ~1 hour
>
> Can you confirm that we're talking about the 32-bit multilib of libgcc here?
> Then, if the answer was "no", we need to understand why. The check is
> simple:
>
> dnl Check if the compiler is configured for setjmp/longjmp exceptions.
> AC_DEFUN([GCC_CHECK_SJLJ_EXCEPTIONS],
> [AC_CACHE_CHECK([whether the compiler is configured for setjmp/longjmp
> exceptions],
> ac_cv_sjlj_exceptions,
> [AC_COMPILE_IFELSE(
> [AC_LANG_PROGRAM(
> [[#ifdef __USING_SJLJ_EXCEPTIONS__
> this will fail
> #endif]],
> [[int i;]])],
> [ac_cv_sjlj_exceptions=no],
> [ac_cv_sjlj_exceptions=yes])])])
>
> Can you run it manually with -m32?
For 64bits the result is "no":
643 configure:4751: checking whether the compiler is configured for
setjmp/longjmp exceptions
644 configure:4769: x86_64-w64-mingw32-gcc
-L/home/andoni/mingw/windows/multilib/x86_64-w64-mingw32/lib
-L/home/andoni/mingw/windows/multilib/mingw/lib -isystem /home
/andoni/mingw/windows/multilib/x86_64-w64-mingw32/include -isystem
/home/andoni/mingw/windows/multilib/mingw/include -c -Wall -g -O2
-DWINVER=0x0600 -D_WIN32_ WINNT=0x0600 conftest.c >&5
645 conftest.c: In function 'main':
646 conftest.c:30:5: warning: unused variable 'i' [-Wunused-variable]
647 int i;
648 ^
649 configure:4769: $? = 0
650 configure:4776: result: no
for 32 bits the result is "yes":
631 configure:4751: checking whether the compiler is configured for
setjmp/longjmp exceptions
632 configure:4769: x86_64-w64-mingw32-gcc
-L/home/andoni/mingw/windows/multilib/x86_64-w64-mingw32/lib
-L/home/andoni/mingw/windows/multilib/mingw/lib -isystem /home
/andoni/mingw/windows/multilib/x86_64-w64-mingw32/include -isystem
/home/andoni/mingw/windows/multilib/mingw/include -m32 -c -Wall -g -O2
-DWINVER=0x0600 -D_WIN 32_WINNT=0x0600 conftest.c >&5
633 conftest.c:25:11: error: unknown type name 'this'
634 this will fail
635 ^~~~
636 conftest.c:25:21: error: expected '=', ',', ';', 'asm' or '__attribute__'
before 'fail'
637 this will fail
638 ^~~~
639 configure:4769: $? = 1
[...]
674 configure:4776: result: yes
For 32 bits the final EH model being used is Dwarf2:
340 # Additional sources to handle exceptions; overridden by targets as needed.
341 LIB2ADDEH = $(srcdir)/unwind-dw2.c $(srcdir)/unwind-dw2-fde.c \
342 $(srcdir)/unwind-sjlj.c $(srcdir)/unwind-c.c
What I understand is happening in a multilib build is that for 64 it chooses
SEH because ac_cv_sjlj_exceptions is not "yes" but for 32, it chooses dw2
instead of SJLJ for the same reason, because ac_cv_sjlj_exceptions is not
"yes":
if test x$ac_cv_sjlj_exceptions = xyes; then
tmake_eh_file="i386/t-sjlj-eh"
elif test "${host_address}" = 32; then
# biarch -m32 with --disable-sjlj-exceptions
tmake_eh_file="i386/t-dw2-eh"
md_unwind_header=i386/w32-unwind.h
else
tmake_eh_file="i386/t-seh-eh"
fi
What should happen in a mutilib build is that if the user configured without
passing any specific options for EH, the 32 bits target should be forced to be
configured with --enable-sjlj-exceptions.
