Thanks for reviewing, Pedro.


Pedro Alves wrote:
> On Wednesday 24 June 2009 19:50:39, Aleksandar Ristovski wrote:
> 
>> Ok, here is my new patch. I addressed all of the above, and 
>> probably introduced some new issues :-).  For my bonus 
>> points, I added comments for each function definition in 
>> nto-low.c
> 
> :-) Thanks for that.
> 
> On Wednesday 24 June 2009 19:50:39, Aleksandar Ristovski wrote:
>> +  /* Skip over argc, argv and envp... (see comment in ldd.c)  */
> 
> Are ldd.c's sources available?  :-)  If not, could you paste
> that comment there too?

Comment from ldd.c pasted.

> 
> On Wednesday 24 June 2009 19:50:39, Aleksandar Ristovski wrote
>> +  i[34567]86-*-nto*)   srv_regobj=reg-i386.o
>> +                       srv_tgtobj="nto-low.o nto-x86-low.o"
>> +                       srv_qnx_LIBS=-lsocket
>> +                       srv_qnx="yes"
> 
> Do you think the set of libs will change depending on the qnx
> arch?  In other words, if you're adding a $srv_qnx, do you really
> need a new srv_qnx_LIBS variable?

Unlikely. I just thought it was nicer. But in this patch I 
removed it and -lsocket is now "hard-coded" in 
configure.ac/configure.

> 
>> +  //TODO: FPU, XMM registers
>> +  return -1;
> 
> Is this a planned feature?  An nto limitation?  Anyway, the reason
> this caught my eye was due to use of C++ style comment.  For
> consistency, please always use C-style comments (yes, even though
> gcc supports those as C extension for ages).

I will be adding this; for now, just removed the comment.

> 
> 
>> Let me know what you think (once this goes in, I will change 
>> gdb's configure.tgt to say "yes" to generating gdbserver for 
>> Neutrino - in a separate patch submission).
>>
> 
> Other than the obvious watchpoint/point interface
> changes that happened meanwhile, and the nits above, it looks OK to me.
> 
> 

Renamed.


Additionally, there are a few small fixes for problems I 
discovered while testing it:
- in find_new_threads, I needed to check if returned 
status.tid is zero (which may happen when program dies).
- in nto_resume added FLTSTACK and FLTACCESS.
- in nto_wait, assign reported signal to 
nto_inferior.exit_signo. Current code in nto-procfs.c 
doesn't do that but it's a bug that results gdb reporting 
"program exited normally" after program dies due to 
unhandled signal.


ChangeLog:
	Adding Neutrino gdbserver.
	* configure: Regenerated.
	* configure.ac: Add case for srv_qnx and set LIBS 
accordingly..
	* configure.srv (i[34567]86-*-nto*): New target.
	* nto-low.c, nto-low.h, nto-x86-low.c: New files.
	* remote-utils.c [__QNX__]: Include sys/iomgr.h
	(nto_comctrl) [__QNX__]: New function.
	(enable_async_io, disable_async_io) [__QNX__]: Call 
nto_comctrl.


-- 
Aleksandar Ristovski
QNX Software Systems