Re: kqueue
> >> On Nov 24, 2016, at 3:27 PM, Alex Zavatone wrote: >> >> >> Yeah, I was trying to see if it would be possible for whatever service is >> failing to send events to display periodic output that would help you >> determine what stops working and why. >> >> I'll send you some samples offline that might help. I received your examples. Mac ones could be useful. Thanks. Although the code does pass all the Basic JDK WatchService tests, so I think my code is pretty much correct. I would still be interested if anyone had any thoughts on why java going into a wait would cause the kqueue to stop posting events? Or would know of general reasons why kqueue would stop posting events? I did a lot of googling on that the last time around but pretty much came up empty. Michael Hall ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
Re: kqueue
On Nov 24, 2016, at 2:40 PM, Michael Hall wrote: >> On Nov 24, 2016, at 2:06 PM, Alex Zavatone wrote: >> >> How long does it take before it stops? Is that time consistent? > > kqueue seems to stop posting events as soon as the java code starts waiting > for it to post them. > The kevent wait looks currently looks like this… > > n = kevent64( queueFD, NULL, 0, &ev, 1, 0, &timeout ); > > With the timeout included this will return control to my code but with no > events posted. > >> >> If it is, what is driving the event broadcasting and what conditions are >> causing it so stop? >> > See above. There are many events pending but they seem to stop posting once > java processes it’s first batch and goes into a wait for more. > >> Is there a way for your event broadcaster to fire off a useless heartbeat >> every 15 seconds so that you can see that it's alive and you can monitor the >> log just for the heartbeat in another window so that you can catch when it >> stops and see what conditions changed to make it stop? > > As above, with no timeout the code never posts. With a timeout it > successfully returns to my code but no events are posted. >> >> A few years ago, when running one app in the iOS Simulator, I used to >> redirect NSLog to a file on my HD that I could tail in the a terminal window. >> >> Then I'd open up n+1 terminal windows, with each one tail-ing the log file. >> I'd use AWK or SED or something to pattern match and filter the output so >> that each window would track specific conditions and output. >> >> This made it convenient to monitor the patient (the app) and keep track of >> specific conditions on a per window (per readout) basis. > > Not sure I’m following. I can output messages fine. But I’m not sure what I > can put in them that will tell me anything since nothing is actually > happening. > > Michael Hall > Yeah, I was trying to see if it would be possible for whatever service is failing to send events to display periodic output that would help you determine what stops working and why. I'll send you some samples offline that might help. Cheers. > > >> >> >> GL. >> >> >> On Nov 24, 2016, at 1:52 PM, Michael Hall wrote: >> >>> This is not entirely Cocoa, it is java, but it involves ObjectiveC JNI, the >>> java native interface. >>> Also a little bit of a long story, my apologies. >>> Specifically the question concerns kqueue, my code is based on the Uli >>> Kusterer ObjectiveC interface to that. >>> Java nio.2 introduced the WatchService class, this monitors directories and >>> tells you of any file system changes. >>> For OS X the implementation is a pure java polling one. Unlike about every >>> other platform where they are native based. >>> I tried to write a native OS X one based on kqueue. >>> I got it so that it passes the nio.2 Basic WatchService tests. >>> There is another set of tests for the code under load. LotsOfEvents. I ran >>> into what seems a deadlock condition between the java code and the kqueue >>> code. >>> Left it at that months if not years ago, but the code was sitting up on >>> github. >>> I noticed someone had recently forked it. I looked at their fork out of >>> curiosity and they said LotOfEvents ran. >>> I figured some minor miracle had occurred. I finally sat down and looked at >>> it today and unfortunately it is just that I left the code in a state where >>> it didn’t do the actual test to see if it was successful, which it is not, >>> but instead just runs through and finishes without indicating any errors. >>> So it just looked like it worked. >>> >>> What the test code does is fire a lot of events. It processes what it has >>> and then waits to see if it gets more. >>> >>> For some reason while it waits for more events kqueue stops posting events >>> entirely. If you check for anything returned the test fails. >>> If you don’t, but let the test to continue to run. kqueue happily resumes >>> pumping events with the debug messages making it look like it works. >>> >>> Any thoughts? I thought I’d ask here before I spend a whole lot of time (or >>> LotsOfTime) spinning wheels back at square 1 again. >>> >>> Michael Hall >>> >>> >>> >>> >>> ___ >>> >>> Cocoa-dev mailing list (Cocoa-de
Re: kqueue
> On Nov 24, 2016, at 2:06 PM, Alex Zavatone wrote: > > How long does it take before it stops? Is that time consistent? kqueue seems to stop posting events as soon as the java code starts waiting for it to post them. The kevent wait looks currently looks like this… n = kevent64( queueFD, NULL, 0, &ev, 1, 0, &timeout ); With the timeout included this will return control to my code but with no events posted. > > If it is, what is driving the event broadcasting and what conditions are > causing it so stop? > See above. There are many events pending but they seem to stop posting once java processes it’s first batch and goes into a wait for more. > Is there a way for your event broadcaster to fire off a useless heartbeat > every 15 seconds so that you can see that it's alive and you can monitor the > log just for the heartbeat in another window so that you can catch when it > stops and see what conditions changed to make it stop? As above, with no timeout the code never posts. With a timeout it successfully returns to my code but no events are posted. > > A few years ago, when running one app in the iOS Simulator, I used to > redirect NSLog to a file on my HD that I could tail in the a terminal window. > > Then I'd open up n+1 terminal windows, with each one tail-ing the log file. > I'd use AWK or SED or something to pattern match and filter the output so > that each window would track specific conditions and output. > > This made it convenient to monitor the patient (the app) and keep track of > specific conditions on a per window (per readout) basis. Not sure I’m following. I can output messages fine. But I’m not sure what I can put in them that will tell me anything since nothing is actually happening. Michael Hall > > > GL. > > > On Nov 24, 2016, at 1:52 PM, Michael Hall wrote: > >> This is not entirely Cocoa, it is java, but it involves ObjectiveC JNI, the >> java native interface. >> Also a little bit of a long story, my apologies. >> Specifically the question concerns kqueue, my code is based on the Uli >> Kusterer ObjectiveC interface to that. >> Java nio.2 introduced the WatchService class, this monitors directories and >> tells you of any file system changes. >> For OS X the implementation is a pure java polling one. Unlike about every >> other platform where they are native based. >> I tried to write a native OS X one based on kqueue. >> I got it so that it passes the nio.2 Basic WatchService tests. >> There is another set of tests for the code under load. LotsOfEvents. I ran >> into what seems a deadlock condition between the java code and the kqueue >> code. >> Left it at that months if not years ago, but the code was sitting up on >> github. >> I noticed someone had recently forked it. I looked at their fork out of >> curiosity and they said LotOfEvents ran. >> I figured some minor miracle had occurred. I finally sat down and looked at >> it today and unfortunately it is just that I left the code in a state where >> it didn’t do the actual test to see if it was successful, which it is not, >> but instead just runs through and finishes without indicating any errors. So >> it just looked like it worked. >> >> What the test code does is fire a lot of events. It processes what it has >> and then waits to see if it gets more. >> >> For some reason while it waits for more events kqueue stops posting events >> entirely. If you check for anything returned the test fails. >> If you don’t, but let the test to continue to run. kqueue happily resumes >> pumping events with the debug messages making it look like it works. >> >> Any thoughts? I thought I’d ask here before I spend a whole lot of time (or >> LotsOfTime) spinning wheels back at square 1 again. >> >> Michael Hall >> >> >> >> >> ___ >> >> Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) >> >> Please do not post admin requests or moderator comments to the list. >> Contact the moderators at cocoa-dev-admins(at)lists.apple.com >> >> Help/Unsubscribe/Update your Subscription: >> https://lists.apple.com/mailman/options/cocoa-dev/zav%40mac.com >> >> This email sent to z...@mac.com > ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
Re: kqueue
How long does it take before it stops? Is that time consistent? If it is, what is driving the event broadcasting and what conditions are causing it so stop? Is there a way for your event broadcaster to fire off a useless heartbeat every 15 seconds so that you can see that it's alive and you can monitor the log just for the heartbeat in another window so that you can catch when it stops and see what conditions changed to make it stop? A few years ago, when running one app in the iOS Simulator, I used to redirect NSLog to a file on my HD that I could tail in the a terminal window. Then I'd open up n+1 terminal windows, with each one tail-ing the log file. I'd use AWK or SED or something to pattern match and filter the output so that each window would track specific conditions and output. This made it convenient to monitor the patient (the app) and keep track of specific conditions on a per window (per readout) basis. GL. On Nov 24, 2016, at 1:52 PM, Michael Hall wrote: > This is not entirely Cocoa, it is java, but it involves ObjectiveC JNI, the > java native interface. > Also a little bit of a long story, my apologies. > Specifically the question concerns kqueue, my code is based on the Uli > Kusterer ObjectiveC interface to that. > Java nio.2 introduced the WatchService class, this monitors directories and > tells you of any file system changes. > For OS X the implementation is a pure java polling one. Unlike about every > other platform where they are native based. > I tried to write a native OS X one based on kqueue. > I got it so that it passes the nio.2 Basic WatchService tests. > There is another set of tests for the code under load. LotsOfEvents. I ran > into what seems a deadlock condition between the java code and the kqueue > code. > Left it at that months if not years ago, but the code was sitting up on > github. > I noticed someone had recently forked it. I looked at their fork out of > curiosity and they said LotOfEvents ran. > I figured some minor miracle had occurred. I finally sat down and looked at > it today and unfortunately it is just that I left the code in a state where > it didn’t do the actual test to see if it was successful, which it is not, > but instead just runs through and finishes without indicating any errors. So > it just looked like it worked. > > What the test code does is fire a lot of events. It processes what it has and > then waits to see if it gets more. > > For some reason while it waits for more events kqueue stops posting events > entirely. If you check for anything returned the test fails. > If you don’t, but let the test to continue to run. kqueue happily resumes > pumping events with the debug messages making it look like it works. > > Any thoughts? I thought I’d ask here before I spend a whole lot of time (or > LotsOfTime) spinning wheels back at square 1 again. > > Michael Hall > > > > > ___ > > Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) > > Please do not post admin requests or moderator comments to the list. > Contact the moderators at cocoa-dev-admins(at)lists.apple.com > > Help/Unsubscribe/Update your Subscription: > https://lists.apple.com/mailman/options/cocoa-dev/zav%40mac.com > > This email sent to z...@mac.com ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
kqueue
This is not entirely Cocoa, it is java, but it involves ObjectiveC JNI, the java native interface. Also a little bit of a long story, my apologies. Specifically the question concerns kqueue, my code is based on the Uli Kusterer ObjectiveC interface to that. Java nio.2 introduced the WatchService class, this monitors directories and tells you of any file system changes. For OS X the implementation is a pure java polling one. Unlike about every other platform where they are native based. I tried to write a native OS X one based on kqueue. I got it so that it passes the nio.2 Basic WatchService tests. There is another set of tests for the code under load. LotsOfEvents. I ran into what seems a deadlock condition between the java code and the kqueue code. Left it at that months if not years ago, but the code was sitting up on github. I noticed someone had recently forked it. I looked at their fork out of curiosity and they said LotOfEvents ran. I figured some minor miracle had occurred. I finally sat down and looked at it today and unfortunately it is just that I left the code in a state where it didn’t do the actual test to see if it was successful, which it is not, but instead just runs through and finishes without indicating any errors. So it just looked like it worked. What the test code does is fire a lot of events. It processes what it has and then waits to see if it gets more. For some reason while it waits for more events kqueue stops posting events entirely. If you check for anything returned the test fails. If you don’t, but let the test to continue to run. kqueue happily resumes pumping events with the debug messages making it look like it works. Any thoughts? I thought I’d ask here before I spend a whole lot of time (or LotsOfTime) spinning wheels back at square 1 again. Michael Hall ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
Re: kqueue and kevent
On Sun, Mar 13, 2011 at 4:48 PM, Jeffrey Walton wrote: > Be careful of this sort of functionality, especially when an adversary > controls the event. In essence, the event should be consider > "untrusted user input". Unless I'm missing something, I don't see how this is any more dangerous than calling select(2). I mean, yes, if someone writes to a pipe, and you read from it, then you'll read what they wrote. The same caution applies to any external data: sockets, files on disk, shared memory… --Kyle Sluder ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
Re: kqueue and kevent
On Sun, Mar 13, 2011 at 3:15 PM, Dave Keck wrote: >> I have used kqueue and kevents for event triggering. However i am not sure >> if it is possible to send events to a kqueue? Googling didnt helped. I was >> thinking as its "kernel" que and kernel notifies. Does it mean that users >> can not send events >> to a queue other than signals? > > The kernel is the only entity that can generate events. You can easily > cause an event to be generated though by, for example, writing to a > pipe for which the respective read end has a EVFILT_READ filter > monitoring it. > Be careful of this sort of functionality, especially when an adversary controls the event. In essence, the event should be consider "untrusted user input". Jeff ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
Re: kqueue and kevent
> I have used kqueue and kevents for event triggering. However i am not sure > if it is possible to send events to a kqueue? Googling didnt helped. I was > thinking as its "kernel" que and kernel notifies. Does it mean that users > can not send events > to a queue other than signals? The kernel is the only entity that can generate events. You can easily cause an event to be generated though by, for example, writing to a pipe for which the respective read end has a EVFILT_READ filter monitoring it. BTW, this question is better suited for darwin-dev. ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
kqueue and kevent
Hi, I have used kqueue and kevents for event triggering. However i am not sure if it is possible to send events to a kqueue? Googling didnt helped. I was thinking as its "kernel" que and kernel notifies. Does it mean that users can not send events to a queue other than signals? Abhi ___ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
