> : manu@trotsky</usr/local/jdk13/bin>./java -green -version
> : java: ../../../../../src/solaris/hpi/green_threads/src/dl-malloc.c:1636:
> : malloc_extend_top: Assertion `((size_t)((char*)(((mbinptr)(&(av_[2 *
> : (0)])))->fd) + top_size) & (pagesz - 1)) == 0' failed.
> : Abort (core dumped)
> 
> Then something is borking internal structures of the JVM; possibly, we don't
> emulate what's needed to do Linux longjmp()/siglongjmp() properly.

I tried this test program:

#include <setjmp.h>
#include <stdio.h>

void g(void);
jmp_buf env;
int i = 0;

int main (int argc, char* argv) {
        if (setjmp(env) != 0) {
                 (void) printf("2nd return from setjmp: i = %d\n", i);
                exit(0);
        }
        (void)  printf("1st return from setjmp: i = %d\n", i);
        i = 1;
        g();
        /*NOTREACHED*/
}

void g(void) {
        longjmp(env, 1);
        /*NOTREACHED*/
}

It works as expected in emulation:
1st return from setjmp:  i = 0
2nd return from setjmp: i = 1

But I don't understand what support should be necessary here? Aren't
setjmp/longjmp library only calls (I mean they do not cause system
call)?

-- 
Emmanuel Dreyfus.  
p99dreyf@criens.u-psud.fr