[patch] Increase the bar of printing `handling possible serial event'

[patch] Increase the bar of printing `handling possible serial event'

[Yao Qi]

Hi,
When I debug gdbserver by setting debug_thread to 1, I get many lines
of `handling possible serial event' in log, which dose not carry any
useful information for debugging purpose.

This patch is to increase the threshold of printing this sentence.  Or we
may remove this print completely, because I can't figure out a case
some one wants to read many lines of `handling possible serial event'
in log.

-- 
Yao (齐尧) 

	* server.c (handle_serial_event): Increase the bar to print
	debug message.

diff --git a/gdb/gdbserver/server.c b/gdb/gdbserver/server.c
index 4612457..845ae3a 100644
--- a/gdb/gdbserver/server.c
+++ b/gdb/gdbserver/server.c
@@ -3217,7 +3217,7 @@ process_serial_event (void)
 int
 handle_serial_event (int err, gdb_client_data client_data)
 {
-  if (debug_threads)
+  if (debug_threads > 3)
     fprintf (stderr, "handling possible serial event\n");
 
   /* Really handle it.  */

Re: [patch] Increase the bar of printing `handling possible serial event'

[Pedro Alves]

On Thursday 10 November 2011 06:27:34, Yao Qi wrote:
> Hi,
> When I debug gdbserver by setting debug_thread to 1, I get many lines
> of `handling possible serial event' in log, which dose not carry any
> useful information for debugging purpose.
> 
> This patch is to increase the threshold of printing this sentence.  Or we
> may remove this print completely, because I can't figure out a case
> some one wants to read many lines of `handling possible serial event'
> in log.

I've found it more than once very useful when debugging non-stop/async mode
issues, where the order of serial and target events isn't predictable,
and knowing when a packet was just handled can help figure out
following weird behaviors.

If it bothers so much, I could go with outputting it only when non_stop is
on.  Just bumping is not good, because you can't set debug_thread to anything
other than 1 or 0 without recompiling gdbserver, and that's not good when
asking users for logs.

-- 
Pedro Alves

Re: [patch] Increase the bar of printing `handling possible serial event'

[Stan Shebs]

On 11/10/11 11:41 AM, Pedro Alves wrote:
> On Thursday 10 November 2011 06:27:34, Yao Qi wrote:
>> Hi,
>> When I debug gdbserver by setting debug_thread to 1, I get many lines
>> of `handling possible serial event' in log, which dose not carry any
>> useful information for debugging purpose.
>>
>> This patch is to increase the threshold of printing this sentence.  Or we
>> may remove this print completely, because I can't figure out a case
>> some one wants to read many lines of `handling possible serial event'
>> in log.
> I've found it more than once very useful when debugging non-stop/async mode
> issues, where the order of serial and target events isn't predictable,
> and knowing when a packet was just handled can help figure out
> following weird behaviors.
>
> If it bothers so much, I could go with outputting it only when non_stop is
> on.  Just bumping is not good, because you can't set debug_thread to anything
> other than 1 or 0 without recompiling gdbserver, and that's not good when
> asking users for logs.
>
--debug=2 ?

Might as well invest, the poor code is getting plenty of use/abuse. :-)

Stan

Re: [patch] Increase the bar of printing `handling possible serial event'

[Yao Qi]

On 11/11/2011 03:41 AM, Pedro Alves wrote:
> I've found it more than once very useful when debugging non-stop/async mode
> issues, where the order of serial and target events isn't predictable,
> and knowing when a packet was just handled can help figure out
> following weird behaviors.
> 

OK, I understand.

> If it bothers so much, I could go with outputting it only when non_stop is

No.  "--debug=N" is a right way to go in long term.

-- 
Yao (齐尧)