Steve Sabljak wrote:
The CGI daemon still dies with a SIGSEGV on Solaris 8, worker mpm. Here's a
backtrace.

 ffffffff7e100ae8 memset (0, 0, 504f5354202f6368, ffffffffffffffc0,
504f5354202f6340, 0) + 114
 ffffffff7c404784 get_req (9, 10020eaa0, ffffffff7ffff810, ffffffff7ffff808,
...
And here is a memory dump of the cgid_req_t passed to get_req (100 bytes)

0xffffffff7ffff770:      0x00000001 0x00000000 0x00000000 0x000003c6
0xffffffff7ffff780:      0x00000000 0x00000000 0x00000000 0x00000000
0xffffffff7ffff790:      0x00000000 0x00000000 0x00000000 0x00000052
0xffffffff7ffff7a0:      0x00000000 0x0000002e 0x00000000 0x0000000c
0xffffffff7ffff7b0:      0x00000000 0x00000022 0x00000000 0x00000000
0xffffffff7ffff7c0:      0x00000000 0x00000000 0x00000004 0x00000000
0xffffffff7ffff7d0:      0x00000001 0x00210238 0x00000001 0x0020ed40
0xffffffff7ffff7e0:      0x00000000 0x00000000 0x00000001 0x0020eaa0
0xffffffff7ffff7f0:      0x00000001 0x00000004 0x00000004 0x00000004

I don't see anything in the cgid_req_t which would cause us to apr_pcalloc() (and thus memset()) 0x504f5354202f6368 bytes. Any chance you can do a build with symbols (CFLAGS=-g or --enable-maintainer-mode) and use a debugger that will display line numbers for the backtrace?


Does this happen for any CGI?

Thanks,

Jeff



Reply via email to