On 14 May 2015 at 14:58, Jeff Law <law@redhat.com> wrote:
On 05/13/2015 02:51 AM, Iain Buclaw wrote:
If a symbol that has so far been valid abruptly ends then we will want
to fail the process rather than silently succeed.
---
libiberty/ChangeLog
2015-05-13 Iain Buclaw <ibuclaw@gdcproject.org>
* d-demangle.c (dlang_call_convention): Return NULL if have reached
the
end of the symbol, but expected more.
(dlang_attributes): Likewise.
(dlang_function_type): Likewise.
(dlang_type): Likewise.
(dlang_identifier): Likewise.
(dlang_value): Likewise.
I spot checked various callers of these functions that not return NULL and
they looked reasonable. Though I was a bit concerned about the callers of
dlang_type, dlang_value and dlang_identifier.
In those cases we'll often still do the string_append, string_setlength and
other calls in the caller of dlang_{value,type,identifier}. I'm assuming
that's safe (the error still appears to be bubbling up properly).
The string routines should be safe for that (appending a string with a
zero length does nothing, for instance).
If you can confirm that we're OK in those cases, then this is OK for the
trunk.
I can start fuzzing mangle strings to test that failures actually fail
gracefully. There's already a fuzzer that exists for C++, I think the
only change that would be required for D is exchanging "_Z" for "_D"
and calling cplus_demangle with DMGL_DLANG.