diff options
author | Jason Molenda <jmolenda@apple.com> | 1998-12-04 04:46:18 +0000 |
---|---|---|
committer | Jason Molenda <jmolenda@apple.com> | 1998-12-04 04:46:18 +0000 |
commit | edc452d07d89e2e7f5ce7e8ea77a46d3d4a8d529 (patch) | |
tree | 2547e447a6b5007bd0ba29b96bb8aa78d6a6f42f /gdb/monitor.c | |
parent | 33ccdb1b9786497b5c284c28520888166da877fa (diff) | |
download | binutils-gdb-edc452d07d89e2e7f5ce7e8ea77a46d3d4a8d529.tar.gz |
1998-12-03 Jason Molenda (jsm@bugshack.cygnus.com)
* monitor.c (monitor_read_memory): Zero out pattern buffers
before calling re_search.
(parse_register_dump): Ditto.
PR 18049. This bug had existed erratically since I upgraded to
the new gnu-regex.c this last summer. The problem is mostly in
parse_register_dump; the allocated structure has some random values
in it and there is a flag set in the register_pattern structure by the
gnu-regex library which indicates that the values in the re_registers
should be trusted.
If those arbitrary contents aren't zero, gnu-regex tries to run realloc
on them and we get a core dump on some hosts for some targets when the
moon is just right.
Diffstat (limited to 'gdb/monitor.c')
-rw-r--r-- | gdb/monitor.c | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/gdb/monitor.c b/gdb/monitor.c index 60cfc143e71..ddd24c25a94 100644 --- a/gdb/monitor.c +++ b/gdb/monitor.c @@ -847,6 +847,8 @@ parse_register_dump (buf, len) points to the start of the register value. */ struct re_registers register_strings; + memset (®ister_strings, 0, sizeof (struct re_registers)); + if (re_search (®ister_pattern, buf, len, 0, len, ®ister_strings) == -1) break; @@ -1759,6 +1761,7 @@ monitor_read_memory (memaddr, myaddr, len) struct re_registers resp_strings; RDEBUG(("MON getmem.resp_delim %s\n",current_monitor->getmem.resp_delim)) ; + memset (&resp_strings, 0, sizeof (struct re_registers)); tmp = strlen (p); retval = re_search (&getmem_resp_delim_pattern, p, tmp, 0, tmp, &resp_strings); |