question, xmalloc and gdbserver?

question, xmalloc and gdbserver?

[Michael Snyder]

What's the policy about using malloc vs. xmalloc in gdbserver?
I notice both are used, with xmalloc being favored, but there are
still quite a few calls to malloc.

Re: question, xmalloc and gdbserver?

[Pedro Alves]

On Tuesday 01 March 2011 23:25:27, Michael Snyder wrote:
> What's the policy about using malloc vs. xmalloc in gdbserver?
> I notice both are used, with xmalloc being favored, but there are
> still quite a few calls to malloc.

I think malloc is used when a failure is not to be considered
fatal, e.g., when the size of the data is determined by the input.

Doug went through and fixed a bunch a while ago (yearly 2009).
You should be able to find the discussions around the patches
in the archives.  Basically, if there's no explicit check
for NULL after malloc, then it's quite likely a new call
that went in after Doug's fixes.

-- 
Pedro Alves

Re: question, xmalloc and gdbserver?

[Michael Snyder]

Pedro Alves wrote:
> On Tuesday 01 March 2011 23:25:27, Michael Snyder wrote:
>> What's the policy about using malloc vs. xmalloc in gdbserver?
>> I notice both are used, with xmalloc being favored, but there are
>> still quite a few calls to malloc.
> 
> I think malloc is used when a failure is not to be considered
> fatal, e.g., when the size of the data is determined by the input.
> 
> Doug went through and fixed a bunch a while ago (yearly 2009).
> You should be able to find the discussions around the patches
> in the archives.  Basically, if there's no explicit check
> for NULL after malloc, then it's quite likely a new call
> that went in after Doug's fixes.
> 

OK, so you're saying that all "legitimate" calls to malloc
will be checked?

Re: question, xmalloc and gdbserver?

[Pedro Alves]

On Wednesday 02 March 2011 18:31:28, Michael Snyder wrote:
> OK, so you're saying that all "legitimate" calls to malloc
> will be checked?

Yes.

-- 
Pedro Alves

Re: question, xmalloc and gdbserver?

[Michael Snyder]

Pedro Alves wrote:
> On Wednesday 02 March 2011 18:31:28, Michael Snyder wrote:
>> OK, so you're saying that all "legitimate" calls to malloc
>> will be checked?
> 
> Yes.
> 

Well, here's the one I'm thinking of.  It's not clear to me
what to do in case of error.  What do you think?

/* Put a stop reply to the stop reply queue.  */

static void
queue_stop_reply (ptid_t ptid, struct target_waitstatus *status)
{
   struct vstop_notif *new_notif;

   new_notif = malloc (sizeof (*new_notif));
   new_notif->next = NULL;
   new_notif->ptid = ptid;
   new_notif->status = *status;

[...]