Building libsigsegv on Cygwin64
Ken Brown
kbrown@cornell.edu
Sun Mar 2 17:21:00 GMT 2014
On 2/21/2014 10:32 AM, Angelo Graziosi wrote:
> Trying to build libsigsegv-2.10 on Cygwin64 (using the src tarball and
> its .cygport file from x86 distribution), fails as follows
>
> [...]
> libtool: compile: gcc -DHAVE_CONFIG_H -I.
> -I/works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src -I.. -I.
> -I/works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src -ggdb -O2 -pipe
> -Wimplicit-function-declaration
> -fdebug-prefix-map=/works/tmp/libsigsegv-2.10-1/build=/usr/src/debug/libsigsegv-2.10-1
> -fdebug-prefix-map=/works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10=/usr/src/debug/libsigsegv-2.10-1
> -c /works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src/handler.c
> -DDLL_EXPORT -DPIC -o .libs/handler.o
> In file included from
> /works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src/handler.c:20:0:
> /works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src/handler-win32.c: In
> function 'main_exception_filter':
> /works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src/handler-win32.c:218:43:
> error: 'struct _CONTEXT' has no member named 'Esp'
> ExceptionInfo->ContextRecord->Esp = new_safe_esp;
> ^
> /works/tmp/libsigsegv-2.10-1/src/libsigsegv-2.10/src/handler-win32.c:220:43:
> error: 'struct _CONTEXT' has no member named 'Eip'
> ExceptionInfo->ContextRecord->Eip = (unsigned
> long)&stack_overflow_handler;
> ^
> Makefile:399: set di istruzioni per l'obiettivo "handler.lo" non riuscito
> make[1]: *** [handler.lo] Errore 1
> make[1]: uscita dalla directory "/works/tmp/libsigsegv-2.10-1/build/src"
> Makefile:344: set di istruzioni per l'obiettivo "install-recursive" non
> riuscito
> make: *** [install-recursive] Errore 1
>
> Since my Cygwin64 is a fesh installation, I wonder if I missed to
> installe some needed packages... or is that error to be expected on
> Cygwin64?
I found the problem (or at least I found *a* problem): There's a
configure test "checking whether a fault handler according to POSIX
works", which passes on 32-bit Cygwin but fails on 64-bit Cygwin. I'm
attaching a file containing the configure test. Here's what happens in
the 64-bit case:
$ gcc -o fault fault.c
$ ./fault.exe
$ echo $?
1
In the 32-bit case, the exit code is 0.
I don't know if this indicates a Cygwin bug or something wrong with the
test.
Ken
-------------- next part --------------
#include <stdlib.h>
#include <signal.h>
#include <sys/signal.h>
#include <sys/types.h>
#include <sys/mman.h>
# include <fcntl.h>
# define zero_fd -1
# define map_flags MAP_ANON | MAP_PRIVATE
# define SIGSEGV_FAULT_ADDRESS_ROUNDOFF_BITS 0UL
unsigned long page;
int handler_called = 0;
void sigsegv_handler (int sig, siginfo_t *sip, void *ucp)
{
void *fault_address = (void *) (sip->si_addr);
handler_called++;
if (handler_called == 10)
exit (4);
if (fault_address
!= (void*)((page + 0x678) & ~SIGSEGV_FAULT_ADDRESS_ROUNDOFF_BITS))
exit (3);
if (mprotect ((void *) page, 0x10000, PROT_READ | PROT_WRITE) < 0)
exit (2);
}
void crasher (unsigned long p)
{
*(int *) (p + 0x678) = 42;
}
int main ()
{
void *p;
struct sigaction action;
/* Preparations. */
/* Setup some mmaped memory. */
p = mmap ((void *) 0x12340000, 0x10000, PROT_READ | PROT_WRITE, map_flags, zero_fd, 0);
if (p == (void *)(-1))
exit (2);
page = (unsigned long) p;
/* Make it read-only. */
if (mprotect ((void *) page, 0x10000, PROT_READ) < 0)
exit (2);
/* Install the SIGSEGV handler. */
sigemptyset(&action.sa_mask);
action.sa_sigaction = &sigsegv_handler;
action.sa_flags = SA_SIGINFO;
sigaction (SIGSEGV, &action, (struct sigaction *) NULL);
sigaction (SIGBUS, &action, (struct sigaction *) NULL);
/* The first write access should invoke the handler and then complete. */
crasher (page);
/* The second write access should not invoke the handler. */
crasher (page);
/* Check that the handler was called only once. */
if (handler_called != 1)
exit (1);
/* Test passed! */
return 0;
}
-------------- next part --------------
--
Problem reports: http://cygwin.com/problems.html
FAQ: http://cygwin.com/faq/
Documentation: http://cygwin.com/docs.html
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
More information about the Cygwin
mailing list