I shall try and test SAT on KF.
Aggelos Mpimpoudis <[EMAIL PROTECTED]> Sent by: [EMAIL PROTECTED] 08/02/2007 09:22 AM Please respond to OSGi Developer Mail List <[email protected]> To OSGi Developer Mail List <[email protected]> cc Subject Re: [osgi-dev] Service Tracking Matter Mr Simon, Is there any possibility to bring SAT, into extensive testing mode, for bug hunting purposes? Aggelos Mpimpoudis Peter Kriens wrote: Stop! This is an OSGi-dev list and saying that your OSGi Toolkit is not running on any Framework but Equinox is the wrong way out imho. :-) Watching this thread I think somebody from the SAT team should take a look what goes wrong because it is likely that most of the debugging will be in the SAT code. That is, it is easier for you prove Knopflerfish wrong because you got the spec than Knopflerfish proving your code wrong. If you advertise something as an OSGi toolkit I think the onus is on you to test it on the popular frameworks. Simon, can't you run the tutorial yourself on Knopflerfish and see what is going on? BTW, it would be really nice if people included source code in these bundles so you can more easily debug them. Just place the source code in OPT-INF/src (bnd can do this automatically) and Eclipse will automagically discover it. Kind regards, Peter Kriens SJA> SJA> Aggelos SJA> SJA> Thanks for persevering! You've done a good job.... my feeling SJA> is that this is a bug in Knopflerfish, although as I've said SJA> before, I've not run any tests myself. SAT relies on OSGi's SJA> ServiceListener/ServiceEvent mechanism, which it uses to track SJA> the availability of services. It does not use the SJA> ServiceTracker, since it pre-dates it by about a year. The last SJA> time I looked, the ServiceTracker is also implemented using the SJA> ServiceListener/ServiceEvent mechanism. SJA> SJA> While you are welcome to try using the ServiceTracker on SJA> Knopflerfish (as an alternative to SAT), my concern would be that SJA> this might not actually fix anything IF the problem is in the SJA> Knopflerfish framework implementation. My gut tells me that the SJA> problem is related to Knopflerfish's firing of ServiceEvents. SJA> SJA> My recommendation to you are two-told: SJA> SJA> 1. Contact either a committer from the Knopflerfish project, or SJA> someone that has experience with Knopflerfish. Use them to help SJA> identify if this is indeed a bug in the Knopflerfish framework implementation. SJA> SJA> 2. Consider switching to Equinox. I am, of course, biased SJA> towards Equinox, having used it from the very start. SJA> SJA> SJA> While I understand that switching to Equinox is not ideal for SJA> you and your team, it seems to be the most productive path. Not SJA> only is Equinox well supported by the Eclipse IDE (unsurprising, SJA> I know), but it is also the OSGi reference implementation for the core spec. SJA> SJA> Good luck, SJA> SJA> Simon SJA> SJA> SJA> SJA> Aggelos Mpimpoudis <[EMAIL PROTECTED]> SJA> Sent by: [EMAIL PROTECTED] SJA> 08/01/2007 08:18 PM SJA> SJA> Please respond to SJA> OSGi Developer Mail List <[email protected]> SJA> SJA> SJA> To SJA> OSGi Developer Mail List <[email protected]> SJA> cc SJA> SJA> Subject SJA> Re: [osgi-dev] Service Tracking Matter SJA> SJA> SJA> SJA> SJA> I bundlized manually the classes. Well here is the KF result. SJA> SJA> 1. Loaded all the appropriate Bundles. SJA> 2. Stopped all of them to emulate an empty framework with SJA> installed bundles...and here it is SJA> 3. Started Ventor Service (nothing really happed because it is just a package) SJA> 4. Started Popcorn. Activated, ok! SJA> 5. Started HotDog. Activated, ok! SJA> 6. Started Customer. Started, Activated, and consumed a product from a ventor, ok! SJA> 7. Now I deactivate HotDog. Deactivated, ok! SJA> 8. I also deactivate PopCorn. Customer & PopCorn Deactivated, ok! SJA> 9. PopCorn Re-Activated........Customer is sleeping, strike 1! SJA> 10. Hotdog Re-Activated........Customer is still sleeping, strike 2! SJA> So, as you see, there is something wrong going on with SAT on SJA> KF. On the contrary, doing the same on Equinox reproduces the SJA> normal sequence (of course, as it was designed and tested for equinox). SJA> I provide attached to this mail, the used bundles and the SJA> relevant screenshots as an informal proof :) SJA> SJA> So, my options are limited to 3: I switch to equinox (very SJA> difficult, because it will affect nearly 15 ppl at the research SJA> group) I switch to service tracking....really...not in the mood, SJA> after I saw what SAT can do Or last but not least, wait, if there SJA> is a chance to track and fix this, in order to provide SJA> functionality also for knopflerfish (if yes pls let me know). I SJA> will be glad to have SAT-enabled bundles into our project. If the SJA> problem is in knopflerfish itself, it would be nice to give SJA> feedback at the KS's developing group, for a bug fix. SJA> If I say anything wrong, feel free to comment on it. SJA> Thank you very much for your cooperation at this and at SJA> previous threads. I would like very much to continue our talk, SJA> provided that the osgi list operators won't find the subject more SAT-oriented and less OSGi :-) SJA> SJA> My very best regards, SJA> Aggelos Mpimpoudis SJA> SJA> SJA> SJA> Simon J Archer wrote: SJA> Please try, at least a simple test case, on Equinox + SAT. I SJA> am wondering whether the bug is in Knopflerfish too, but I would SJA> not want to blame anyone without having run some tests myself. SJA> The SAT tutorial includes a zip that contains all the code you SJA> need to simply exercise the tutorial without working it by SJA> hand.... this might be a quick way for you to confirm Equinox vs. SJA> KF, and to confirm your understanding. SJA> SJA> Remember, SAT is simply a bundle built using the OSGi APIs... SJA> it has no dependency upon any particular implementation. SJA> SJA> I am happy to help... how do you want to proceed? SJA> SJA> Simon SJA> SJA> -- Dept. of Informatics & Telecommunications, University of Athens Athens, Greece Gsm: +306942075153 / Skype: aggelos.mpimpoudis email: a.mpimpoydhs [at] di.uoa.gr _______________________________________________ OSGi Developer Mail List [email protected] http://www2.osgi.org/mailman/listinfo/osgi-dev
_______________________________________________ OSGi Developer Mail List [email protected] http://www2.osgi.org/mailman/listinfo/osgi-dev
