flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Harui <aha...@adobe.com>
Subject Re: AW: AW: [VMs] Monitoring volunteer(s) needed
Date Wed, 23 Jul 2014 17:52:47 GMT
I'm fine using some other kill technology. Java's is apparently not good enough.
Sent via the PANTECH Discover, an AT&T 4G LTE smartphone.

Christofer Dutz <christofer.dutz@c-ware.de> wrote:


Eventually we shouldn't implement our own killer, but let the OS do that for us.
Would something like this eventually help? We could implement a simple if(os-a) {} else if
(os-b)  {} else if(os-x) {} thing with platform dependent kill commands that we execute.
http://www.tech-recipes.com/rx/446/xp_kill_windows_process_command_line_taskkill/

Chris
________________________________________
Von: Alex Harui <aharui@adobe.com>
Gesendet: Mittwoch, 23. Juli 2014 08:51
An: dev@flex.apache.org
Betreff: Re: AW: [VMs] Monitoring volunteer(s) needed

Runner.java is the Mustella test runner.  From my playing around with it,
it also has a bunch of timers looking for test output on a socket.  There
are timers for how long it should take for a process to start up and one
watching for test steps to complete and some others.  But a stuck player
somehow doesn't always get killed.  Not sure why.  IIRC, once the run-time
exception dialog box is up, Windows won't always kill it.

-Alex

On 7/22/14 11:46 PM, "Christofer Dutz" <christofer.dutz@c-ware.de> wrote:

>How does the Mustella testrunner actually work?
>
>In flexmojos we grab a socket and compile that socket number into the
>test-swf. The test swfs contain test-listeners depending on the unit-test
>framework used that connect to that socket and report back to the server
>how the tests went. Here whenever something is received we reset a timer.
>If that timer expires the process is killed. I think that would be quite
>good for the Mustella tests too as it doesn't measure the overall test
>duration, but how long a single test takes. That should work for the
>DataGrid tests (At least I watched about 100 different tests running
>there in one Flashplayer instance).
>
>Chris
>________________________________________
>Von: Alex Harui <aharui@adobe.com>
>Gesendet: Mittwoch, 23. Juli 2014 08:11
>An: dev@flex.apache.org
>Betreff: Re: [VMs] Monitoring volunteer(s) needed
>
>On 7/22/14 2:49 PM, "Michael A. Labriola" <labriola@digitalprimates.net>
>wrote:
>
>>>As far as I understood the problem. Whenever the build was hanging, I
>>>could see a Flashplayer instance complaining about a timeout of 15
>>>seconds having expired and if the user whiches to continue or cancel,
>>>hereby blocking the instance and the associated test.
>>
>>I never finished it but we were playing with a watch dog for FlexUnit so
>>that the test runner sent a heartbeat message back to a monitoring
>>instance (the java process that launched it) every (n) frames. Then we
>>could configure the amount of time it was okay to go without the
>>heartbeat. The difference being we would wait so long as Flash Player was
>>still well enough to send its heartbeat. When it ceased, we would kill
>>it.
>In the mustella code is a .c source code for KillWin, something I wrote
>for Windows that used to be able to kill off the player.  I haven't tried
>on recent players.  But I haven't seen stuck player on the builds machine.
> We're having different issues there.
>
>-Alex
>


Mime
View raw message