[Laszlo-dev] How to profile application operation

[Henry Minsky]

Hmm, well maybe my lossage had to do with me invoking the Profiler
start/stop using a mouse click handler. That might have caused some kind of
race condition in the IE7 engine, between the profiler task running on the
timer and a manual event calling Profiler.stop. I'll use your tool to avoid
invoking the profiler via the mouse.

On Thu, May 7, 2009 at 1:46 PM, Henry Minsky <henry.minsky at gmail.com> wrote:

> No it works in Firefox on OSX, something is screwing up in IE7. WHen I look
> at the data packets, there is something odd. The chunks have plausible data,
> but then one of the chunks just has "</xml>" in it, and that is not the
> last sequence number of data that gets sent (i.e., it is seqnum=3, and then
> there's more data in seqnum=4 packet)..
>
>
> On Thu, May 7, 2009 at 1:43 PM, P T Withington <ptw at pobox.com> wrote:
>
>> Where are you trying this?  Because the code that I wrote below worked
>> fine for me.
>>
>> Is it possible your changes to move the URL construction out of the
>> profiler somehow affected the ability to restart the profiler?
>>
>>
>> On 2009-05-07, at 13:39EDT, Henry Minsky wrote:
>>
>>  I am having difficulty getting the profiler to start up again. If I load
>>> the
>>> main.lzx app with profiling on,and then call Profiler.start  [I had a
>>> clickable view to invoke Profiler.start()  , I am not getting any
>>> new output to the main.profiler file.
>>>
>>> Even in a simple test case, like this
>>>
>>> <canvas>
>>>   <view width="40" height="40" bgcolor="red" onclick="Profiler.stop();
>>> alert('stop profiler');"/>
>>>   <view width="40" height="40"  bgcolor="green" onclick="alert('start
>>> profiler'); Profiler.start();"/>
>>> </canvas>
>>>
>>> The new profiler file does not get written out. In a trace of HTTP
>>> request
>>> via Fiddler, I see all
>>> the profiler data chunks go out, and the close packet is sent.  Just for
>>> some reason the new profiler
>>> output file is not getting written on the server. WTF?
>>>
>>>
>>>
>>> On Thu, May 7, 2009 at 12:52 PM, P T Withington <ptw at pobox.com> wrote:
>>>
>>>  Here's a hack that I wrote that waits for the instantiator to settle
>>>> out,
>>>> then profiles for a bit.  You could use this to make a profile of a
>>>> running
>>>> application use case:
>>>>
>>>> <text id="myoutput" multiline="true" />
>>>>
>>>> <handler name="oninited">
>>>>  LzTimeKernel.setTimeout(function () {
>>>>    if (lz.Instantiator.isUpdating) {
>>>>      // If the instantiator is still running, just requeue ourselves
>>>>      LzTimeKernel.setTimeout(arguments.callee, 10 * 1000);
>>>>    } else {
>>>>      // Wait a bit for things to settle
>>>>      LzTimeKernel.setTimeout(function () {
>>>>        myoutput.bringToFront();
>>>>        myoutput.format("LzIdleKernel.__callbacks: %w\n",
>>>> LzIdleKernel.__callbacks);
>>>>        myoutput.addFormat("lz.Idle.coi: %w\n", lz.Idle.coi);
>>>>        if ($profile) {
>>>>          Profiler.stop();
>>>>          myoutput.addFormat("Starting profiling...");
>>>>          Profiler.start();
>>>>          // Profile for 40 seconds
>>>>          LzTimeKernel.setTimeout(function () {
>>>>            Profiler.stop();
>>>>            myoutput.addFormat(" done!\n");
>>>>          }, 40 * 1000);
>>>>        }
>>>>      }, 10 * 1000);
>>>>    }
>>>>  }, 10 * 1000);
>>>> </handler>
>>>>
>>>>
>>>>
>>>
>>> --
>>> Henry Minsky
>>> Software Architect
>>> hminsky at laszlosystems.com
>>>
>>
>>
>
>
> --
> Henry Minsky
> Software Architect
> hminsky at laszlosystems.com
>
>
>


-- 
Henry Minsky
Software Architect
hminsky at laszlosystems.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.openlaszlo.org/pipermail/laszlo-dev/attachments/20090507/3c5a535d/attachment.html