loading MarvinSketch/View 5.10.2 in MacOS 10.6, Safari 5.1.7

I've recently upgraded to MacOS 10.6, so I upgraded ACE to deliver Marvin 5.4.1.1 when the user is using MacOS 10.5, and Marvin 5.10.2 when the user is using MacOS 10.6 or higher or any other OS.  

I'm finding that both MarvinView and MarvinSketch are very slow to load the first time I go to a page that uses them, and they will often crash the third or fourth time that I go to a page that uses them.  After they crash, I can reload or go to a new page, and the whole process starts again.

I don't see anything appear in the Java console.  It crashes when MarvinSketch or MarvinView does, and restarts when they do.  

I am having similar problems with Chrome.

Any advice?

Hi Bob,

yes, unfortunatelly the new features and improvements increase the size of the applet, but besides this in the 5.10 series there are some jars that are loading unexpectedly. We are working on to fix this loading issues and reduce the initial size of the applet in 5.10. The fix should be included in a patch release called 5.10.4 or 5.10.5, it is planned to be released in the middle of september (for 5.10.4 we have a fixed latest commit date, and currently we are not sure we can fix this issues until that time). Unfortunatelly I could not really suggest any workaround for now.

Are there any error log that is available regarding to the crashes, and can they be also reproduced on other environments? Are the applet examples from the chemaxon.com server also crash? Can I test somehow with your site to see a crash on one of our MacOSX?

Regards,
Istvan

I don't see anything in the Java console.  I can usually get Marvin to reload by navigating away from the page that displays Marvin and then navigating back, so no permanent damage is done, but it's a real pain in the butt, especially when it takes a long time to decide to crash.  

Even before Marvin crashes, navigating away from a page sometimes takes a long time to enact.  It appears that the MarvinView applet takes a long time to shut itself down.

You can go to http://epoch.uky.edu/ace/public/parity.jsp and submit responses repeatedly.  This page does not use AJAX, so it mimics the feel of navigating to different pages each time you submit a structure.

Firefox seems to work better than either Safari or Chrome, BTW.

Here's another observation that may help: If I navigate away from a page that contains a lot of MarvinView instances, it takes the browser a fair amount of time to close all those instances before it will navigate away from them.

Hi Bob,

finally I could set up the same platform you use, now I can reproduce your Mac related issues, and can find some answers for you. Sorry about the late silence, but it was a bit hard to get a working 10.6.8 since we had not have one before, and we also were a bit overwhelmed unfortunatelly.

For this reload issue, I have found some information in the log and it seems it has relations with some problems with the HearBeat that controls whether the applet is still running, and ensure to kill the applet some time after the destroy call.

The HeartBeat mechanism had several issues, and I found some memory related (http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6788906) and Javascript communication related (http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6995583) bugs in the bug database, based on this, I have tried to increase the heap size under the applet, and I could reproduce the issue significantly rare with the parity page. So this may be a problem in our side and the applet have some leakage, but after I have added -Xmx1024m to the JVM parameters, it has taken more then 40 reload to reproduce the issue. I am filing a ticket into our tracking sytem about the possible leakage, and we will investigate this in the future.

Regards,
Istvan

Thanks, Istvan.  I started having hard drive problems on my MacOS 10.6 system, so I got another new computer and I am now up to 10.7.5, but I am still seeing the same issues, not just on Safari but in other browsers as well, though they seem particularly severe in Safari.

I concur with your assessment that the symptoms are consistent with problems in closing the applet.  

The sooner you can fix this problem, the better.  It is extremely aggravating.  A lot of the students are using Macs these days, as well, so it is not a problem just for the instructors.  

Have you made progress in fixing these problems in Marvin 5.11?  It seems to load a little better than previously, but it's hard to tell.

Hi Bob!

   Unfortunatelly we haven't got time investigate this possible leakage, however it might got better by a sideeffect of an other change. It is still on table for us, but not for our 5.11 branch. I hope you can accept this answer.

Regards:
Balázs

bzaicsek wrote:

Unfortunatelly we haven't got time investigate this possible leakage, however it might got better by a sideeffect of an other change. It is still on table for us, but not for our 5.11 branch. I hope you can accept this answer.

I guess I don't have much of a choice, do I?  

I hope you will work on it for the 5.12 branch.  Remember, a LOT of the students (and instructors) are using Macs these days.

Dear Bob!

   We have examined your problem, and has some good and bad news.

• Freezing applet on tabchange:
  
  
  
  • we found a workaround for the problem, it is not the best solution, but solves things for now. (It will be in our next release.)
  
  
  
  


• Slow applet strat:
  
  
  
  • we have tried to make it better
  
  
  
  


• Slow applet resume:
  
  
  
  • we have found the root of the problem, but haven't got time to make it better
  
  
  
  

   Please be patient, and wait for our next release (5.12.0). Until that: could you test your applets with Opera? Opera on OS X handles Java and Java Applets as other browsers on Windows, which can solve some of your problems (at least, it has hidden in our tests), but unfortunatelly it is not fully compatible with Java on Mac OS, so it can become unstable.

   At this time we could only do that, but we are eager to solve the identifed issue.

Regards:
Balázs 

Thanks for the update.  Do you have a planned release date for 5.12?

Sorry for the late answer.  The current planned date is 10th of February.