Re: [Valgrind-developers] valgrind: r9397 - in branches/DARWIN: coregrind coregrind/m_scheduler coregrind/m_sigframe coregrind/m_syswrap include/vki memcheck/tests

[Nicholas N. <n.n...@gm...>]

On Sat, Mar 14, 2009 at 7:10 AM, Nicholas Nethercote
<n.n...@gm...> wrote:
> On Sat, Mar 14, 2009 at 1:40 AM,  <sv...@va...> wrote:
>>
>> - If a thread is waiting in a syscall, and is interrupted by a
>>  signal, its state is not restored properly.  (at all?)
>>
>> - Guest syscalls are run without setting the correct signal mask.
>>  (this point and previous requires on fixing syscall-x86-darwin.S)
>>
>> - the signal delivery frames are completely different from the
>>  real kernel ones.  Consequently any client code that relies
>>  on the precise layout of these frames will fail (stack
>>  unwinders).
>>
>> - siginfo and ucontext parameters (2nd and 3rd params to signal
>>  handlers) are not implemented; NULL is passed for both.
>>  Handlers which inspect these will segfault.
>
> Is it worth documenting these points in a comment somewhere?

ps: this is a great commit, I get 40 fewer test failures now (266 -> 226).

N