[v3] Introduce remote_target_is_gdbserver

  This patch introduces a function in gdbserver-support.exp to find out
whether the current target is GDBserver.

The code was inspired from gdb.trace/qtor.exp, so it replaces the code
there by a call to the new function.

New in v3:
- Remove useless "pass" in remote_target_is_gdbserver.
- Coding style in qtro.exp (braces in condition).
- Changelog entry about qtro.exp.

gdb/testsuite/ChangeLog:

	* gdb.trace/qtro.exp: Replace gdbserver detection code by...
	* lib/gdbserver-support.exp (remote_target_is_gdbserver): New
	function.
---
 gdb/testsuite/gdb.trace/qtro.exp        | 14 +-------------
 gdb/testsuite/lib/gdbserver-support.exp | 18 ++++++++++++++++++
 2 files changed, 19 insertions(+), 13 deletions(-)
  

Simon Marchi <simon.marchi@ericsson.com> writes:
> This patch introduces a function in gdbserver-support.exp to find out
> whether the current target is GDBserver.
>
> The code was inspired from gdb.trace/qtor.exp, so it replaces the code
> there by a call to the new function.
>
> New in v3:
> - Remove useless "pass" in remote_target_is_gdbserver.
> - Coding style in qtro.exp (braces in condition).
> - Changelog entry about qtro.exp.
>
> gdb/testsuite/ChangeLog:
>
> 	* gdb.trace/qtro.exp: Replace gdbserver detection code by...
> 	* lib/gdbserver-support.exp (remote_target_is_gdbserver): New
> 	function.
> ---
>  gdb/testsuite/gdb.trace/qtro.exp        | 14 +-------------
>  gdb/testsuite/lib/gdbserver-support.exp | 18 ++++++++++++++++++
>  2 files changed, 19 insertions(+), 13 deletions(-)
>
> diff --git a/gdb/testsuite/gdb.trace/qtro.exp b/gdb/testsuite/gdb.trace/qtro.exp
> index 22b5051..700c157 100644
> --- a/gdb/testsuite/gdb.trace/qtro.exp
> +++ b/gdb/testsuite/gdb.trace/qtro.exp
> @@ -98,19 +98,7 @@ if { $traceframe_info_supported == -1 } {
>  }
>  
>  # Check whether we're testing with our own GDBserver.
> -set is_gdbserver -1
> -set test "probe for GDBserver"
> -gdb_test_multiple "monitor help" $test {
> -    -re "The following monitor commands are supported.*debug-hw-points.*remote-debug.*GDBserver.*$gdb_prompt $" {
> -	set is_gdbserver 1
> -	pass $test
> -    }
> -    -re "$gdb_prompt $" {
> -	set is_gdbserver 0
> -	pass $test
> -    }
> -}
> -if { $is_gdbserver == -1 } {
> +if { ![remote_target_is_gdbserver] } {
>      return -1
>  }
>  
> diff --git a/gdb/testsuite/lib/gdbserver-support.exp b/gdb/testsuite/lib/gdbserver-support.exp
> index 026a937..423c729 100644
> --- a/gdb/testsuite/lib/gdbserver-support.exp
> +++ b/gdb/testsuite/lib/gdbserver-support.exp
> @@ -436,3 +436,21 @@ proc mi_gdbserver_start_multi { } {
>  
>      return [mi_gdb_target_cmd $gdbserver_protocol $gdbserver_gdbport]
>  }
> +
> +# Return true if the current remote target is an instance of gdbserver.
> +
> +proc remote_target_is_gdbserver { } {
> +    global gdb_prompt
> +
> +    set is_gdbserver 0
> +    set test "Probing for GDBserver"
> +
> +    gdb_test_multiple "monitor help" $test {
> +	-re "The following monitor commands are supported.*Quit GDBserver.*$gdb_prompt $" {
> +		set is_gdbserver 1
> +	}
> +	-re "$gdb_prompt $" {
> +	}
> +    }
> +    return $is_gdbserver
> +}

Hi.

The original code allowed for a -1 value of is_gdbserver
to handle the case of "can't tell" (e.g. for a timeout or
whatever, IIUC).
While IWBN to not complicate the API of
remote_target_is_gdbserver by requiring the caller
to have to handle this, maybe it'd be best if the caller
did have to watch for -1 and not just assume "not gdbserver":
maybe a different test will want to handle all three cases
(can't-tell, no, or yes).
E.g., initialize is_gdbserver to -1, and watch for a -1 value
before returning.

proc remote_target_is_gdbserver { } {
    global gdb_prompt

    set is_gdbserver -1
    set test "Probing for GDBserver"

    gdb_test_multiple "monitor help" $test {
	-re "The following monitor commands are supported.*Quit GDBserver.*$gdb_prompt $" {
	    set is_gdbserver 1
	}
	-re "$gdb_prompt $" {
	    set is_gdbserver 0
	}
    }
    if { $is_gdbserver == -1 } {
	verbose -log "can't tell if using gdbserver or not" # or whatever
	set $is_gdbserver 0  # <<<< this part I'm not sure about
    }
    return $is_gdbserver

Also, I see an earlier version of the patch first
did a check for [is_remote_target] before calling
target_is_gdbserver, and the new version of the
patch changes that to just calling remote_target_is_gdbserver.
Since the function remote_target_is_gdbserver can
be used regardless of whether the target is remote,
let's remove "remote_" from the name.
ie., go back to target_is_gdbserver.

Hmmm, another thought.
Since this requires an exchange with the target,
IWBN to cache the result.
There's support for doing this in the harness,
grep for gdb_caching_proc.
  

On 09/11/2014 11:54 AM, Simon Marchi wrote:
> This patch introduces a function in gdbserver-support.exp to find out
> whether the current target is GDBserver.
>
> The code was inspired from gdb.trace/qtor.exp, so it replaces the code
> there by a call to the new function.
>
> New in v3:
> - Remove useless "pass" in remote_target_is_gdbserver.
> - Coding style in qtro.exp (braces in condition).
> - Changelog entry about qtro.exp.
>
> gdb/testsuite/ChangeLog:
>
> 	* gdb.trace/qtro.exp: Replace gdbserver detection code by...
> 	* lib/gdbserver-support.exp (remote_target_is_gdbserver): New
> 	function.
> ---
>   gdb/testsuite/gdb.trace/qtro.exp        | 14 +-------------
>   gdb/testsuite/lib/gdbserver-support.exp | 18 ++++++++++++++++++
>   2 files changed, 19 insertions(+), 13 deletions(-)
>
> diff --git a/gdb/testsuite/gdb.trace/qtro.exp b/gdb/testsuite/gdb.trace/qtro.exp
> index 22b5051..700c157 100644
> --- a/gdb/testsuite/gdb.trace/qtro.exp
> +++ b/gdb/testsuite/gdb.trace/qtro.exp
> @@ -98,19 +98,7 @@ if { $traceframe_info_supported == -1 } {
>   }
>
>   # Check whether we're testing with our own GDBserver.
> -set is_gdbserver -1
> -set test "probe for GDBserver"
> -gdb_test_multiple "monitor help" $test {
> -    -re "The following monitor commands are supported.*debug-hw-points.*remote-debug.*GDBserver.*$gdb_prompt $" {
> -	set is_gdbserver 1
> -	pass $test
> -    }
> -    -re "$gdb_prompt $" {
> -	set is_gdbserver 0
> -	pass $test
> -    }
> -}
> -if { $is_gdbserver == -1 } {
> +if { ![remote_target_is_gdbserver] } {
>       return -1
>   }
>
> diff --git a/gdb/testsuite/lib/gdbserver-support.exp b/gdb/testsuite/lib/gdbserver-support.exp
> index 026a937..423c729 100644
> --- a/gdb/testsuite/lib/gdbserver-support.exp
> +++ b/gdb/testsuite/lib/gdbserver-support.exp
> @@ -436,3 +436,21 @@ proc mi_gdbserver_start_multi { } {
>
>       return [mi_gdb_target_cmd $gdbserver_protocol $gdbserver_gdbport]
>   }
> +
> +# Return true if the current remote target is an instance of gdbserver.
> +
> +proc remote_target_is_gdbserver { } {
> +    global gdb_prompt
> +
> +    set is_gdbserver 0
> +    set test "Probing for GDBserver"
> +
> +    gdb_test_multiple "monitor help" $test {
> +	-re "The following monitor commands are supported.*Quit GDBserver.*$gdb_prompt $" {
> +		set is_gdbserver 1
> +	}
> +	-re "$gdb_prompt $" {
> +	}
> +    }
> +    return $is_gdbserver
> +}
>

Just a thought but... is there a specific reason we need to check for 
gdbserver explicitly as opposed to checking for a remote debugging stub 
feature?

It sounds more appropriate to leave remote stub-based testing as generic 
as possible, in which case "is_gdbserver" wouldn't be needed.
  

On 14-09-11 11:56 AM, Doug Evans wrote:
> Simon Marchi <simon.marchi@ericsson.com> writes:
>> This patch introduces a function in gdbserver-support.exp to find out
>> whether the current target is GDBserver.
>>
>> The code was inspired from gdb.trace/qtor.exp, so it replaces the code
>> there by a call to the new function.
>>
>> New in v3:
>> - Remove useless "pass" in remote_target_is_gdbserver.
>> - Coding style in qtro.exp (braces in condition).
>> - Changelog entry about qtro.exp.
>>
>> gdb/testsuite/ChangeLog:
>>
>> 	* gdb.trace/qtro.exp: Replace gdbserver detection code by...
>> 	* lib/gdbserver-support.exp (remote_target_is_gdbserver): New
>> 	function.
>> ---
>>  gdb/testsuite/gdb.trace/qtro.exp        | 14 +-------------
>>  gdb/testsuite/lib/gdbserver-support.exp | 18 ++++++++++++++++++
>>  2 files changed, 19 insertions(+), 13 deletions(-)
>>
>> diff --git a/gdb/testsuite/gdb.trace/qtro.exp b/gdb/testsuite/gdb.trace/qtro.exp
>> index 22b5051..700c157 100644
>> --- a/gdb/testsuite/gdb.trace/qtro.exp
>> +++ b/gdb/testsuite/gdb.trace/qtro.exp
>> @@ -98,19 +98,7 @@ if { $traceframe_info_supported == -1 } {
>>  }
>>  
>>  # Check whether we're testing with our own GDBserver.
>> -set is_gdbserver -1
>> -set test "probe for GDBserver"
>> -gdb_test_multiple "monitor help" $test {
>> -    -re "The following monitor commands are supported.*debug-hw-points.*remote-debug.*GDBserver.*$gdb_prompt $" {
>> -	set is_gdbserver 1
>> -	pass $test
>> -    }
>> -    -re "$gdb_prompt $" {
>> -	set is_gdbserver 0
>> -	pass $test
>> -    }
>> -}
>> -if { $is_gdbserver == -1 } {
>> +if { ![remote_target_is_gdbserver] } {
>>      return -1
>>  }
>>  
>> diff --git a/gdb/testsuite/lib/gdbserver-support.exp b/gdb/testsuite/lib/gdbserver-support.exp
>> index 026a937..423c729 100644
>> --- a/gdb/testsuite/lib/gdbserver-support.exp
>> +++ b/gdb/testsuite/lib/gdbserver-support.exp
>> @@ -436,3 +436,21 @@ proc mi_gdbserver_start_multi { } {
>>  
>>      return [mi_gdb_target_cmd $gdbserver_protocol $gdbserver_gdbport]
>>  }
>> +
>> +# Return true if the current remote target is an instance of gdbserver.
>> +
>> +proc remote_target_is_gdbserver { } {
>> +    global gdb_prompt
>> +
>> +    set is_gdbserver 0
>> +    set test "Probing for GDBserver"
>> +
>> +    gdb_test_multiple "monitor help" $test {
>> +	-re "The following monitor commands are supported.*Quit GDBserver.*$gdb_prompt $" {
>> +		set is_gdbserver 1
>> +	}
>> +	-re "$gdb_prompt $" {
>> +	}
>> +    }
>> +    return $is_gdbserver
>> +}
> 
> Hi.
> 
> The original code allowed for a -1 value of is_gdbserver
> to handle the case of "can't tell" (e.g. for a timeout or
> whatever, IIUC).
> While IWBN to not complicate the API of
> remote_target_is_gdbserver by requiring the caller
> to have to handle this, maybe it'd be best if the caller
> did have to watch for -1 and not just assume "not gdbserver":
> maybe a different test will want to handle all three cases
> (can't-tell, no, or yes).
> E.g., initialize is_gdbserver to -1, and watch for a -1 value
> before returning.
> 
> proc remote_target_is_gdbserver { } {
>     global gdb_prompt
> 
>     set is_gdbserver -1
>     set test "Probing for GDBserver"
> 
>     gdb_test_multiple "monitor help" $test {
> 	-re "The following monitor commands are supported.*Quit GDBserver.*$gdb_prompt $" {
> 	    set is_gdbserver 1
> 	}
> 	-re "$gdb_prompt $" {
> 	    set is_gdbserver 0
> 	}
>     }
>     if { $is_gdbserver == -1 } {
> 	verbose -log "can't tell if using gdbserver or not" # or whatever
> 	set $is_gdbserver 0  # <<<< this part I'm not sure about
>     }
>     return $is_gdbserver

I am confused. Do you want remote_target_is_gdbserver to return -1 in case of error, or not. The paragraph seems to say yes, but the code seems to say no.

> Also, I see an earlier version of the patch first
> did a check for [is_remote_target] before calling
> target_is_gdbserver, and the new version of the
> patch changes that to just calling remote_target_is_gdbserver.
> Since the function remote_target_is_gdbserver can
> be used regardless of whether the target is remote,
> let's remove "remote_" from the name.
> ie., go back to target_is_gdbserver.

Indeed, if the name is target_is_gdbserver, it would be clearer that you can call it in any situation, even if you are using native.

However, in that particular case, I removed the [is_remote target] check, since there is an equivalent one earlier in the test.

> Hmmm, another thought.
> Since this requires an exchange with the target,
> IWBN to cache the result.
> There's support for doing this in the harness,
> grep for gdb_caching_proc.

Interesting, I will check that.
  

> Just a thought but... is there a specific reason we need to check for 
> gdbserver explicitly as opposed to checking for a remote debugging stub 
> feature?
> 
> It sounds more appropriate to leave remote stub-based testing as generic 
> as possible, in which case "is_gdbserver" wouldn't be needed.

Good point, but I'd rather not address this in the current patch. I don't
want to change the test itself, but just make the check available to other
tests.

Ultimately, I want to use it in this patch [1] where I really want to check
for GDBserver. I want to KFAIL a test only when using gdbserver, due to a
known bug in gdbserver.

[1] https://sourceware.org/ml/gdb-patches/2014-09/msg00148.html
  

On 09/11/2014 04:33 PM, Simon Marchi wrote:
>> Just a thought but... is there a specific reason we need to check for
>> gdbserver explicitly as opposed to checking for a remote debugging stub
>> feature?
>>
>> It sounds more appropriate to leave remote stub-based testing as generic
>> as possible, in which case "is_gdbserver" wouldn't be needed.
>
> Good point, but I'd rather not address this in the current patch. I don't
> want to change the test itself, but just make the check available to other
> tests.
>

Agreed. This may be something for the future. I don't want to create 
more work for you.

> Ultimately, I want to use it in this patch [1] where I really want to check
> for GDBserver. I want to KFAIL a test only when using gdbserver, due to a
> known bug in gdbserver.
>
> [1] https://sourceware.org/ml/gdb-patches/2014-09/msg00148.html
>

I see. Thanks for the explanation.