On 15/12/2015 07:59, Thomas Huth wrote:
> On 12/12/15 11:39, Paolo Bonzini wrote:
>> "Unimplemented" messages go to stderr, everything else goes to tracepoints
>>
>> Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
>> ---
> [...]
>> diff --git a/linux-user/signal.c b/linux-user/signal.c
>> index 9d62e02..919aa83 100644
>> --- a/linux-user/signal.c
>> +++ b/linux-user/signal.c
> [...]
>> @@ -3972,14 +3974,15 @@ give_sigsegv:
>>
>> long do_sigreturn(CPUOpenRISCState *env)
>> {
>> -
>> - qemu_log("do_sigreturn: not implemented\n");
>> + trace_user_do_sigreturn(env, 0);
>> + fprintf(stderr, "do_sigreturn: not implemented\n");
>> return -TARGET_ENOSYS;
>> }
>>
>> long do_rt_sigreturn(CPUOpenRISCState *env)
>> {
>> - qemu_log("do_rt_sigreturn: not implemented\n");
>> + trace_user_do_rt_sigreturn(env, 0);
>> + fprintf(stderr, "do_rt_sigreturn: not implemented\n");
>> return -TARGET_ENOSYS;
>> }
>
> Why not simply using qemu_log_mask(LOG_UNIMP, ...) instead? Isn't that
> what the LOG_UNIMP flag is good for?
LOG_UNIMP has to be enabled explicitly, because it is guest-triggerable.
Here the error is really fatal, so I decided to use fprintf. In fact,
it would probably be better to abort after printing an error, but I left
the "return" untouched because I didn't want to do too many changes in
the same patch.
Paolo
