[19/40] Fix cp_find_first_component_aux bug

  Valgrind catches an out-of-bounds read here:

 $ gdb ./testsuite/outputs/gdb.cp/method2/method2
 (gdb) start
 [...]
 Temporary breakpoint 1, main (argc=1, argv=0x7fffffffd958) at src/gdb/testsuite/gdb.cp/method2.cc:26
 26        return 0;
 (gdb) b A::operator
 ==26907== Invalid read of size 1
 ==26907==    at 0x75C0AE: cp_find_first_component_aux(char const*, int) (cp-support.c:951)
 ==26907==    by 0x75C084: cp_find_first_component(char const*) (cp-support.c:925)
 ==26907==    by 0x75C3DA: cp_entire_prefix_len(char const*) (cp-support.c:1089)
 ==26907==    by 0x758B16: cp_lookup_symbol_in_namespace(char const*, char const*, block const*, domain_enum_tag, int) (cp-namespace.c:314)
 ==26907==    by 0x75972A: lookup_namespace_scope(language_defn const*, char const*, block const*, domain_enum_tag, char const*, int) (cp-namespace.c:739)
 ==26907==    by 0x7597CB: cp_lookup_symbol_nonlocal(language_defn const*, char const*, block const*, domain_enum_tag) (cp-namespace.c:768)
 ==26907==    by 0x8C1137: lookup_symbol_aux(char const*, block const*, domain_enum_tag, language, field_of_this_result*) (symtab.c:2016)
 ==26907==    by 0x8C098A: lookup_symbol_in_language(char const*, block const*, domain_enum_tag, language, field_of_this_result*) (symtab.c:1824)
 ==26907==    by 0x8C0A04: lookup_symbol(char const*, block const*, domain_enum_tag, field_of_this_result*) (symtab.c:1836)
 ==26907==    by 0x82CBE1: find_label_symbols(linespec_state*, VEC_symbolp*, VEC_symbolp**, char const*) (linespec.c:3390)
 ==26907==    by 0x828FB5: linespec_parse_basic(ls_parser*) (linespec.c:1620)
 ==26907==    by 0x82A78F: parse_linespec(ls_parser*, char const*) (linespec.c:2307)
 ==26907==  Address 0x910f97c is 0 bytes after a block of size 12 alloc'd
 ==26907==    at 0x4C28BF6: malloc (vg_replace_malloc.c:299)
 ==26907==    by 0x74E737: xmalloc (common-utils.c:43)
 ==26907==    by 0x74EAF4: savestring(char const*, unsigned long) (common-utils.c:179)
 ==26907==    by 0x826CEF: copy_token_string(ls_token) (linespec.c:488)
 ==26907==    by 0x828EF6: linespec_parse_basic(ls_parser*) (linespec.c:1599)
 ==26907==    by 0x82A78F: parse_linespec(ls_parser*, char const*) (linespec.c:2307)
 ==26907==    by 0x82AE27: event_location_to_sals(ls_parser*, event_location const*) (linespec.c:2469)
 ==26907==    by 0x82B1CE: decode_line_full(event_location const*, int, program_space*, symtab*, int, linespec_result*, char const*, char const*) (linespec.c:2557)
 ==26907==    by 0x720C8A: parse_breakpoint_sals(event_location const*, linespec_result*) (breakpoint.c:9550)
 ==26907==    by 0x72A2F7: create_sals_from_location_default(event_location const*, linespec_result*, bptype) (breakpoint.c:14484)
 ==26907==    by 0x727F86: bkpt_create_sals_from_location(event_location const*, linespec_result*, bptype) (breakpoint.c:13219)
 ==26907==    by 0x72146D: create_breakpoint(gdbarch*, event_location const*, char*, int, char*, int, int, bptype, int, auto_boolean, breakpoint_ops const*, int, int, int, unsigned int) (breakpoint.c:9759)

gdb/ChangeLog:
yyyy-mm-dd  Pedro Alves  <palves@redhat.com>

	* cp-support.c (cp_find_first_component_aux): Add missing case for
	end of string.
---
 gdb/cp-support.c | 2 ++
 1 file changed, 2 insertions(+)
  

On 06/02/2017 05:22 AM, Pedro Alves wrote:
> gdb/ChangeLog:
> yyyy-mm-dd  Pedro Alves  <palves@redhat.com>
> 
> 	* cp-support.c (cp_find_first_component_aux): Add missing case for
> 	end of string.

That's pretty obvious, even to me. Do we a test case would add anything? /me not so sure [but then it would be rather trivial to write, eh?]

Keith
  

On 07/17/2017 12:50 PM, Pedro Alves wrote:
> Would you be OK with adding the fix without the test yet?

Sure.

Keith
  

On 07/17/2017 10:38 PM, Keith Seitz wrote:
> On 07/17/2017 12:50 PM, Pedro Alves wrote:
>> Would you be OK with adding the fix without the test yet?
> 
> Sure.

OK, I pushed it in now.

Thanks,
Pedro Alves