Re: [pldc] Mozart 2.0 issue with Application.exit, and inability to kill ozwish ?

[Sébastien Doeraene]

Hi,

This is a bug. Application.exit should kill ozwish. I don't know about the crash case, I'm not sure there is a clean way to do that anyway.

For the normal exit case, the problem is that we use a brutal std::exit(code);
https://github.com/mozart/mozart2/blob/master/vm/vm/main/modules/modsystem.hh#L159

a cleaner exit should be implemented.

Cheers,

Sébastien

On Thu, Jul 25, 2013 at 5:32 PM, Jérémie Melchior <jeremie....@uclouvain.be> wrote:

Dear all,

I don't know who I should send this e-mail to. I am currently using a lot Mozart 2.0 and I have found some weird behavior (compared to Mozart 1.4).

1°) There seems to be something wrong with the statement {Application.exit 0}. Usually it would close the application and kill the QTk windows opened.

However it does pretty nothing.. except killing the main process (which is the main goal). I guess this is because of the second problem.

2°) When you exit an ozemulator process (with ctrl+c) it does kill the attached processes (ozwish)... good ! But that's not the case when the application crashes or when the application is closed "normally". 

2-a°) in Mozart 1.4 when it crashed, usually the attached processes was still running (normal behavior) however it was easy to kill this process (ctrl+c after the application crash would then kill all the attached processes). This is not the case anymore. I have to type "killall ozwish" everytime the application crash.. that's annoying.

2-b°) in Mozart 1.4 when the application is closed "normally", all the attached processes were killed. That's not the case anymore. (see problem 1°).

Do someone have any idea why the behavior has changed and if it has become a new "normal" behavior or if this should be later fixed ?

Best regards,

Jérémie