Not a problem in karaf. But it looks like IGNORE_INTERRUPS is useless.
For interests I tried a few scenarios to inspect the behavior
The following scenario will fail with full lost of cmd output of each following
command:
System.out.println("5: Wait for Ctrl-C - generateRsaKey
and MThread.sleepInLoop");
long start = System.currentTimeMillis();
while (!MPeriod.isTimeOut(start, sleep)) {
MBouncy.generateRsaKey(MBouncy.RSA_KEY_SIZE.B2048);
}
System.out.println("Finish loop");
if (Thread.interrupted())
throw new InterruptedException();
Thread.sleep(100);
It throws an exception after "Finish loop" but it looks like there is no output
anymore:
karaf@reactive()> shityo stress ctrl-c scenario=5
--- Thread ID 63
>>> Loop 3
5: Wait for Ctrl-C - generateRsaKey and MThread.sleepInLoop
Finish loop
>>> Loop 2
5: Wait for Ctrl-C - generateRsaKey and MThread.sleepInLoop
Finish loop
>>> Loop 1
5: Wait for Ctrl-C - generateRsaKey and MThread.sleepInLoop
^C~
karaf@reactive()> Finish loop
karaf@reactive()> list
karaf@reactive()>
Even for each command!
In the debugger I see that the exception is throw and thats the end of output
to the console from a command.
You can see the full code here:
https://github.com/mhus/mhus-osgi-tools/blob/master/karaf-commands/src/main/java/de/mhus/karaf/commands/testit/StressShit.java
<https://github.com/mhus/mhus-osgi-tools/blob/master/karaf-commands/src/main/java/de/mhus/karaf/commands/testit/StressShit.java>
line 87
But it looks like this is another topic. But is it karaf or felix?
> On 15. Mar 2020, at 06:03, Jean-Baptiste Onofre <[email protected]> wrote:
>
> Hi Mike,
>
> Yes, IMHO there’s no problem on Karaf (sorry for the delay in my answer, the
> situation is "complicated" here ;) ): it just does the interrupt correctly.
>
> It mainly depends how you deal with interrupt in your piece of code.
>
> I’m happy it works now.
>
> Regards
> JB
>
>> Le 15 mars 2020 à 00:22, Mike Hummel <[email protected] <mailto:[email protected]>> a
>> écrit :
>>
>> Hello Guillaume,
>>
>> thank you for the hint. I already tried Thread.interrupted() but found my
>> problem in flag clearing if another wait() is called.
>>
>> To be sure to interrupt a loop I have to check Thread.interrupted() before
>> each Thread.sleep() in the loop. Like this
>>
>> public static void sleepInLoop(long _millisec) throws
>> InterruptedException {
>> if (Thread.interrupted())
>> throw new InterruptedException();
>> Thread.sleep(_millisec);
>> }
>>
>> And this will solfe the problem!
>>
>> Thx,
>>
>> Mike
>>
>>
>>
>>> On 14. Mar 2020, at 21:51, Guillaume Nodet <[email protected]
>>> <mailto:[email protected]>> wrote:
>>>
>>> The Ctrl+C default behavior in Karaf is to to call Thread.interrupt().
>>> The effect depends on what the thread is doing, you'll find more
>>> information at [1].
>>> If you want the command to be interrupted, the code needs to support it
>>> correctly.
>>>
>>> [1]
>>> https://docs.oracle.com/javase/tutorial/essential/concurrency/interrupt.html
>>>
>>> <https://docs.oracle.com/javase/tutorial/essential/concurrency/interrupt.html>
>>>
>>> Cheers,
>>> Guillaume Nodet
>>>
>>>
>>> Le sam. 14 mars 2020 à 20:45, Mike Hummel <[email protected]
>>> <mailto:[email protected]>> a écrit :
>>> The problem is if I not listen to ctrl-c in the moment of the interrupt
>>>
>>> while(true) {
>>> 1) do something
>>> 2) Thread.sleep()
>>> }
>>>
>>> And the 'Interrupt' is while (1) and not in (2) then the cmd will not be
>>> interrupted. - I tested it.
>>>
>>> (The interrupt will be thrown in Thread.sleep() and (I think) in
>>> Object.wait(), but not in 'normal' executon or if try {} (catch Throwable)
>>> is used)
>>>
>>> It will not break the command loop but will move the command in background
>>> (correct: status == Done) and the cmd will not stop working.
>>>
>>> something like this would be nice
>>>
>>> while(true) {
>>> 1) something
>>> 2) Thread.sleep()
>>> 3) if (is canceled) break;
>>> }
>>>
>>> But (in 3) for the command it's not possible to check the status of
>>> execution. It's in org.apache.felix.gogo.runtime.CommandSessionImpl
>>>
>>> So using crtl-c to cancel the command execution is like roulette.
>>>
>>>
>>>
>>>> On 14. Mar 2020, at 20:28, Jean-Baptiste Onofre <[email protected]
>>>> <mailto:[email protected]>> wrote:
>>>>
>>>> CTRL-C just works fine in Karaf commands (including the one creating
>>>> thread like log:tail).
>>>>
>>>> So, your use case is about intercepting CTRL-C yourself, right ? What’s
>>>> the use case you want to achieve ?
>>>>
>>>> Regards
>>>> JB
>>>>
>>>>> Le 14 mars 2020 à 18:26, Mike Hummel <[email protected] <mailto:[email protected]>>
>>>>> a écrit :
>>>>>
>>>>> Looks like it's not possible to block the ctrl-c event. It's implemented
>>>>> in felix gogo shell ans since every cmd (in a pipe) is executed in a
>>>>> separate thread it's not clear witch one should do the interruption
>>>>> control
>>>>>
>>>>> Maybe the feature "IGNORE_INTERRUPS" should not be offered.
>>>>>
>>>>> But felix should inform the commands if the execution is canceled. For
>>>>> example by checking for the Closeable interface and calling the close()
>>>>> method.
>>>>>
>>>>>
>>>>>> On 14. Mar 2020, at 11:09, Jean-Baptiste Onofre <[email protected]
>>>>>> <mailto:[email protected]>> wrote:
>>>>>>
>>>>>> Hi Mike,
>>>>>>
>>>>>> Let me take a look.
>>>>>>
>>>>>> Thanks for the report.
>>>>>>
>>>>>> Regards
>>>>>> JB
>>>>>>
>>>>>>> Le 14 mars 2020 à 09:03, Mike Hummel <[email protected]
>>>>>>> <mailto:[email protected]>> a écrit :
>>>>>>>
>>>>>>> Hello,
>>>>>>>
>>>>>>> I did a little bit research for IGNORE_INTERRUPTS in the karaf sources.
>>>>>>> I found in master and karaf-4.2.x the same results.
>>>>>>>
>>>>>>> The const is defined in 'Sessio'n and 'SessionProperties', but only
>>>>>>> used from 'Session'.
>>>>>>>
>>>>>>> It is only used to enable the behaviour but never used to implement
>>>>>>> some kind of behaviour.
>>>>>>>
>>>>>>> Is there someone how can can prove this? And what is the background for
>>>>>>> two definitions of the same const? If I have the background I could
>>>>>>> look for a fix.
>>>>>>>
>>>>>>> Just now I created KARAF-6645 to track the problem.
>>>>>>>
>>>>>>> Best Regards,
>>>>>>>
>>>>>>> Mike
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> On 5. Mar 2020, at 09:53, Mike Hummel <[email protected]
>>>>>>>> <mailto:[email protected]>> wrote:
>>>>>>>>
>>>>>>>> It's not easy to say ... In 4.2.3 it's broken and in 4.1.x it was ok.
>>>>>>>>
>>>>>>>> I testet the flag IGNORE_INTERRUPTS in 4.2.7 and it was not working as
>>>>>>>> I imagine. I'm even not sure if the flag is exact what I want to do.
>>>>>>>>
>>>>>>>>
>>>>>>>>> On 4. Mar 2020, at 09:18, Guillaume Nodet <[email protected]
>>>>>>>>> <mailto:[email protected]>> wrote:
>>>>>>>>>
>>>>>>>>> Could you be more specific about "older releases" ? Do you know in
>>>>>>>>> which release it broke ?
>>>>>>>>>
>>>>>>>>> Le mer. 4 mars 2020 à 09:14, Mike Hummel <[email protected]
>>>>>>>>> <mailto:[email protected]>> a écrit :
>>>>>>>>> Hello,
>>>>>>>>>
>>>>>>>>> I try to break my karaf commands with Ctrl-C (e.g. by using
>>>>>>>>> Thread.sleep()). In older releases this was no problem, but since the
>>>>>>>>> shell starts every command in a separate thread the Cltr-C is also
>>>>>>>>> caught by gogo and it will unlock the command from the console.
>>>>>>>>>
>>>>>>>>> I also tried this
>>>>>>>>>
>>>>>>>>> session.put(Session.IGNORE_INTERRUPTS, Boolean.TRUE);
>>>>>>>>>
>>>>>>>>> but it gets not the effect.
>>>>>>>>>
>>>>>>>>> Is there a way to recognise if the current command is separated from
>>>>>>>>> the current gogo command line?
>>>>>>>>>
>>>>>>>>> A sample snipped:
>>>>>>>>>
>>>>>>>>> Object oldIgnoreInterrupts =
>>>>>>>>> session.get(Session.IGNORE_INTERRUPTS);
>>>>>>>>> try {
>>>>>>>>> session.put(Session.IGNORE_INTERRUPTS, Boolean.TRUE);
>>>>>>>>> return doExecute(this, cmd, parameters);
>>>>>>>>> } finally {
>>>>>>>>> session.put(Session.IGNORE_INTERRUPTS,
>>>>>>>>> oldIgnoreInterrupts);
>>>>>>>>> }
>>>>>>>>>
>>>>>>>>> public Object doExecute(CmdShitYo base, String cmd, String[]
>>>>>>>>> parameters) throws Exception {
>>>>>>>>> if (cmd.equals("ctrl-c")) {
>>>>>>>>> try {
>>>>>>>>> while (true) {
>>>>>>>>> System.out.println("Wait for Ctrl-C - off");
>>>>>>>>> MThread.sleep(3000);
>>>>>>>>> }
>>>>>>>>> } catch (InterruptedException e) {
>>>>>>>>> System.out.println("Interrupted !!!!");
>>>>>>>>> }
>>>>>>>>> }
>>>>>>>>> }
>>>>>>>>>
>>>>>>>>> MThread:
>>>>>>>>> public static void sleep(long _millisec) {
>>>>>>>>> try {
>>>>>>>>> Thread.sleep(_millisec);
>>>>>>>>> } catch (InterruptedException e) {
>>>>>>>>> log.i(e);
>>>>>>>>> }
>>>>>>>>> }
>>>>>>>>>
>>>>>>>>> This will output
>>>>>>>>>
>>>>>>>>> Wait for Ctrl-C - off
>>>>>>>>> Wait for Ctrl-C - on
>>>>>>>>> Wait for Ctrl-C - off
>>>>>>>>> Wait for Ctrl-C - on
>>>>>>>>> ...
>>>>>>>>>
>>>>>>>>> If I interrupt, it will be separated from gogo shell and iterate for
>>>>>>>>> ever. - And I see the interrupted exception in the log.
>>>>>>>>>
>>>>>>>>> Thx,
>>>>>>>>>
>>>>>>>>> Mike
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> ------------------------
>>>>>>>>> Guillaume Nodet
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>>
>>>
>>> --
>>> ------------------------
>>> Guillaume Nodet
>>>
>>
>
