On 02/17/2012 05:24 PM, Mark Kettenis wrote:
>>
>>  - Check linux/un.h in both GDB and GDBserver's configure script, and
wrap unix socket
>> related code by HAVE_LINUX_UN_H, to make it doesn't break other builds.
>
> I'm pretty sure you should not be including <linux/un.h> directly, but
> use <sys/un.h> instead.

<sys/un.h> is included, and checked in autoconf as well in this new patch.

On 02/24/2012 05:01 AM, Pedro Alves wrote:
>> > +#include <sys/socket.h>
>> > +#include <sys/un.h>
>> >  
> Hard to read, but are these now outside of HAVE_UST?  That'll break
> mingw builds.
> 

They are out of HAVE_UST.  They are warpped by #ifdef IN_PROCESS_AGENT.
 We don't build IN_PROCESS_AGENT on mingw, so it doesn't break mingw build.

I cross build GDB on linux with "--host=i586-mingw32msvc
--target=i586-mingw32msvc", no errors.

>> > @@ -8107,7 +7946,12 @@ gdb_ust_thread (void *arg)
>> >  
>> >  	  if (cmd_buf[0])
>> >  	    {
>> > -	      if (strcmp ("qTfSTM", cmd_buf) == 0)
>> > +	      if (strcmp (cmd_buf, "help") == 0)
>> > +		{
>> > +		  strcpy (cmd_buf, "for help, press F1\n");
>> > +		}
> LOL, I remember writing that as a joke, but I didn't remember
> that it had ended up in the tree.
> 

Removed it for you :)

> Otherwise okay.

This is what I plan to commit once this series are approved totoally.

-- 
Yao (齐尧)