bad news re: mingw exceptions

Thu Feb 28 14:42:00 GMT 2002

Bryce McKinlay <bryce@waitaki.otago.ac.nz> writes:
> > At compile time, how does g++ know that test.cc will get linked
> > against Java code later on (and hence should emit code for Java
> > exceptions, not C++ exceptions)?

> #pragma gcc java_exceptions

Interestingly, when I add this line to test.cc, I am able to link
without -lsupc++. When I do *not* add this line, I must link -lsupc++.
This would seem to indicate that the code I posted earlier was using
C++ exceptions in the CNI case.
However, even when I enable this pragma and do *not* link supc++
(which means that I am *definately* using Java exceptions), the
program runs properly.
> This is declared in some header which is #included from all the CNI 
> headers.

Yes, it's pulled in by javaprims.h, but not by cni.h. Is that how it's
supposed to work?
> If its missing __g++_eh_personality or some such symbol, you're using 

Yeah, that was it, but as I mentioned earlier in this message, it's
moot since even with pragma java_exceptions, exceptions written in C++
work. Very strange, huh?
> My first suspect would be the builtins patch.

Can you give me a pointer to more info on this? I'll roll my source
tree back to before the patch was applied and try again.
> Could you compile your minimal test case with "-O2 -S -fverbose-asm" and 
> post the resulting .s file?

Thanks a lot for taking the time to look at this.
The program below segfaults in _Unwind_SjLj_RaiseException (as a
result of calling _Jv_Throw()). The disassembly for that function
contains only two CALLs, one of which is to a fixed address, and the
other to the value in ecx. The latter call is where it segfaults;
presumably jumping to an invalid location.
 - a
______________________________________________________________________________
Java:
public class test3 {
 public static void main(String[] s) {
 try {
 throw new Exception();
 } catch (Exception e) {
 }
 }
}
______________________________________________________________________________
ASM [gcj -c test3.java -O2 -S -fverbose-asm -o test3.s]
	.file	"test3.java"
 # GNU Java version 3.2 20020227 (experimental) (i686-pc-mingw32)
 #	compiled by GNU C version 2.95.4 20011006 (Debian prerelease).
 # options passed: -fno-use-divide-subroutine -fuse-boehm-gc
 # -fnon-call-exceptions -fkeep-inline-functions -O2 -fverbose-asm
 # options enabled: -fdefer-pop -foptimize-sibling-calls
 # -fcse-follow-jumps -fcse-skip-blocks -fexpensive-optimizations
 # -fthread-jumps -fstrength-reduce -fpeephole -fforce-mem -ffunction-cse
 # -fkeep-inline-functions -finline -fkeep-static-consts -fcaller-saves
 # -freg-struct-return -fgcse -fgcse-lm -fgcse-sm -frerun-cse-after-loop
 # -frerun-loop-opt -fdelete-null-pointer-checks -fschedule-insns2
 # -fsched-interblock -fsched-spec -fbranch-count-reg -fexceptions
 # -funwind-tables -fasynchronous-unwind-tables -fnon-call-exceptions
 # -freorder-blocks -fcprop-registers -fcommon -fverbose-asm -fgnu-linker
 # -fregmove -foptimize-register-move -fargument-alias -fstrict-aliasing
 # -fmerge-constants -fident -fpeephole2 -fguess-branch-probability
 # -fmath-errno -ftrapping-math -fbounds-check -m80387 -mhard-float
 # -mno-soft-float -mieee-fp -mfp-ret-in-387 -mstack-arg-probe
 # -mcpu=pentiumpro -march=i386
	.text
	.align 2
	.p2align 4,,15
.globl __ZN5test3C1Ev
	.def	__ZN5test3C1Ev;	.scl	2;	.type	32;	.endef
__ZN5test3C1Ev:
LFB1:
	pushl	%ebp
LCFI0:
	movl	%esp, %ebp
LCFI1:
	subl	8,ドル %esp
LCFI2:
	movl	%ebp, %esp
	popl	%ebp
	jmp	__ZN4java4lang6ObjectC1Ev
LFE1:
	.def	__Unwind_SjLj_Resume;	.scl	2;	.type	32;	.endef
	.def	___gcj_personality_sj0;	.scl	2;	.type	32;	.endef
	.def	__Unwind_SjLj_Register;	.scl	2;	.type	32;	.endef
	.def	__Unwind_SjLj_Unregister;	.scl	2;	.type	32;	.endef
	.align 2
	.p2align 4,,15
.globl __ZN5test34mainEP6JArrayIPN4java4lang6StringEE
	.def	__ZN5test34mainEP6JArrayIPN4java4lang6StringEE;	.scl	2;	.type	32;	.endef
__ZN5test34mainEP6JArrayIPN4java4lang6StringEE:
LFB2:
	pushl	%ebp
LCFI3:
	movl	%esp, %ebp
LCFI4:
	subl	88,ドル %esp
LCFI5:
	movl	%esp, -16(%ebp)
	leal	-12(%ebp), %eax
	movl	%eax, -24(%ebp)
	leal	-56(%ebp), %eax
	movl	%eax, (%esp)
	movl	%ebx, -12(%ebp)
LCFI6:
	movl	%esi, -8(%ebp)
LCFI7:
	movl	%edi, -4(%ebp)
LCFI8:
	movl	$___gcj_personality_sj0, -32(%ebp)
	movl	$LLSDA0, -28(%ebp)
	movl	$L6, -20(%ebp)
	call	__Unwind_SjLj_Register
	movl	$__ZN5test36class$E, (%esp)
	movl	$-1, -52(%ebp)
	call	__Jv_InitClass
	movl	$__ZN4java4lang9Exception6class$E, (%esp)
	movl	16,ドル 4(%esp)
	movl	1,ドル -52(%ebp)
	call	__Jv_AllocObjectNoFinalizer
	movl	%eax, -60(%ebp)
	movl	%eax, (%esp)
	call	__ZN4java4lang9ExceptionC1Ev
	movl	-60(%ebp), %eax
	movl	%eax, (%esp)
	call	__Jv_Throw
	.p2align 4,,7
L6:
	addl	12,ドル %ebp
	cmpl	1,ドル -44(%ebp)
	movl	-48(%ebp), %eax
	jne	L7
L1:
	movl	$-1, -52(%ebp)
	leal	-56(%ebp), %eax
	movl	%eax, (%esp)
	call	__Unwind_SjLj_Unregister
	movl	-4(%ebp), %edi
	movl	-12(%ebp), %ebx
	movl	-8(%ebp), %esi
	movl	%ebp, %esp
	popl	%ebp
	ret
L7:
	movl	%eax, (%esp)
	movl	$-1, -52(%ebp)
	call	__Unwind_SjLj_Resume
LFE2:
	.section	.gcc_except_table,"w"
	.align 4
LLSDA0:
	.byte	0xff
	.byte	0x0
	.uleb128 LLSDATT0-LLSDATTD0
LLSDATTD0:
	.byte	0x1
	.uleb128 LLSDACSE0-LLSDACSB0
LLSDACSB0:
	.uleb128 0x0
	.uleb128 0x1
LLSDACSE0:
	.byte	0x1
	.byte	0x0
	.align 4
	.long	__ZN4java4lang9Exception6class$E
LLSDATT0:
	.text
	.align 2
__Utf1:
	.word	1465
	.word	4
	.ascii "main"
	.space 1
	.align 2
__Utf2:
	.word	59434
	.word	22
	.ascii "([Ljava.lang.String;)V"
	.space 1
	.align 2
__Utf3:
	.word	626
	.word	6
	.ascii "<init>"
	.space 1
	.align 2
__Utf4:
	.word	39797
	.word	3
	.ascii "()V"
	.space 1
	.data
	.align 32
__MT_test3:
	.long	__Utf1
	.long	__Utf2
	.word	16393
	.word	-1
	.long	__ZN5test34mainEP6JArrayIPN4java4lang6StringEE
	.long	0
	.long	__Utf3
	.long	__Utf4
	.word	16385
	.word	-1
	.long	__ZN5test3C1Ev
	.long	0
.globl __ZTVN5test3E
	.align 32
__ZTVN5test3E:
	.long	0
	.long	0
	.long	__ZN5test36class$E
	.long	8
	.long	__ZN4java4lang6Object8finalizeEv
	.long	__ZN4java4lang6Object8hashCodeEv
	.long	__ZN4java4lang6Object6equalsEPS1_
	.long	__ZN4java4lang6Object8toStringEv
	.long	__ZN4java4lang6Object5cloneEv
	.text
	.align 2
__Utf5:
	.word	19937
	.word	5
	.ascii "test3"
	.space 1
.globl __ZN5test36class$E
	.data
	.align 32
__ZN5test36class$E:
	.long	__ZTVN4java4lang5ClassE+8
	.long	0
	.long	0
	.long	__Utf5
	.word	1
	.space 2
	.long	__ZN4java4lang6Object6class$E
	.long	0
	.long	0
	.long	0
	.long	__MT_test3
	.word	2
	.word	5
	.long	0
	.long	8
	.word	0
	.word	0
	.long	__ZTVN5test3E+8
	.long	0
	.long	0
	.long	0
	.long	0
	.word	0
	.byte	0
	.space 1
	.long	0
	.word	0
	.space 2
	.long	0
	.long	0
	.long	0
	.long	0
	.def	__Jv_RegisterClass;	.scl	2;	.type	32;	.endef
	.text
	.align 2
	.p2align 4,,15
	.def	__GLOBAL__I__ZN5test3C1Ev;	.scl	3;	.type	32;	.endef
__GLOBAL__I__ZN5test3C1Ev:
LFB3:
	pushl	%ebp
LCFI9:
	movl	%esp, %ebp
LCFI10:
	subl	24,ドル %esp
LCFI11:
	movl	$__ZN5test36class$E, (%esp)
	call	__Jv_RegisterClass
	movl	%ebp, %esp
	popl	%ebp
	ret
LFE3:
	.section	.ctors,"w"
	.align 4
	.long	__GLOBAL__I__ZN5test3C1Ev
	.def	__ZN4java4lang6Object5cloneEv;	.scl	3;	.type	32;	.endef
	.def	__ZN4java4lang6Object8hashCodeEv;	.scl	3;	.type	32;	.endef
	.def	__Jv_Throw;	.scl	3;	.type	32;	.endef
	.def	__Jv_AllocObjectNoFinalizer;	.scl	3;	.type	32;	.endef
	.def	__Jv_InitClass;	.scl	3;	.type	32;	.endef
-- 
"If I put copyrighted material into the 'chewy nougat center', I can
 claim the crunchy chocolate coating is an 'Access Control
 Mechanism'." --lynx_user_abroad
Why Windows NT/2k/XP is unreliable: www.zappadoodle.com