ptrace event-loop

ptrace event-loop

[Andrew Cagney]

FYI,

I've added the boolean frysk.sys.Server.useServerXXX, if set to false 
the PtraceServer will be disabled leaving frysk instead using the 
event-loop for all ptrace requests.

Things are at the point where the test results are consistent (actually 
they are slightly better - the hpd tests pass again :-), but I'll leave 
it disabled for a bit longer.

Andrew

Re: ptrace event-loop

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 1126 bytes --]

On Wed, 2007-04-18 at 16:08 -0400, Andrew Cagney wrote:
> FYI,
> 
> I've added the boolean frysk.sys.Server.useServerXXX, if set to false 
> the PtraceServer will be disabled leaving frysk instead using the 
> event-loop for all ptrace requests.
> 
> Things are at the point where the test results are consistent (actually 
> they are slightly better - the hpd tests pass again :-), but I'll leave 
> it disabled for a bit longer.

In CVS useServerXXX = true so it isn't disabled currently by default.
But the results are better than when setting it to false (then the fhpd
tests time out).

With both settings I am seeing TestStepping failures though:

testLineStepFunctionCall(frysk.rt.TestStepping)junit.framework.AssertionFailedError: event loop run explictly stopped (Attempting to add observer)
testLineStepIfStatementFail(frysk.rt.TestStepping)junit.framework.AssertionFailedError: event loop run explictly stopped (Attempting to add observer)
testStepGoto(frysk.rt.TestStepping)junit.framework.AssertionFailedError:
event loop run explictly stopped (Attempting to add observer)

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: ptrace event-loop

[Andrew Cagney]

Mark Wielaard wrote:
> On Wed, 2007-04-18 at 16:08 -0400, Andrew Cagney wrote:
>   
>> FYI,
>>
>> I've added the boolean frysk.sys.Server.useServerXXX, if set to false 
>> the PtraceServer will be disabled leaving frysk instead using the 
>> event-loop for all ptrace requests.
>>
>> Things are at the point where the test results are consistent (actually 
>> they are slightly better - the hpd tests pass again :-), but I'll leave 
>> it disabled for a bit longer.
>>     
>
> In CVS useServerXXX = true so it isn't disabled currently by default.
> But the results are better than when setting it to false (then the fhpd
> tests time out).
>   

Since the Ptrace-Server is enabled, the reliance on the event-loop to 
handle all ptrace calls is _disabled_.

> With both settings I am seeing TestStepping failures though:
>
> testLineStepFunctionCall(frysk.rt.TestStepping)junit.framework.AssertionFailedError: event loop run explictly stopped (Attempting to add observer)
> testLineStepIfStatementFail(frysk.rt.TestStepping)junit.framework.AssertionFailedError: event loop run explictly stopped (Attempting to add observer)
> testStepGoto(frysk.rt.TestStepping)junit.framework.AssertionFailedError:
> event loop run explictly stopped (Attempting to add observer)
>   
There are now four variables to play with:

event-loop \ server-thread:      on    off
   poll
   wait

The TestStepping and other tests show intermittent failures in all 
cases.  We need more narrowly defined and hence more deterministic 
testing here.

Andrew