Rev 3.0 crashing on OS X too (as well as Vista and Linux)

Richard Gaskin ambassador at fourthworld.com
Tue Oct 14 20:00:15 EDT 2008


Bernard Devlin wrote:
> I did try MC with the 3.0 engine on Linux (before finally admitting defeat),
> and I experienced exactly the same kinds of hanging that I experienced with
> the Rev IDE.

Good to know. The solution space just got a little smaller.


> If it was a hardware issue one would expect other applications to suffer.

That would seem logical enough, but may not always be an optimal 
diagnostic measure.

Many years ago I had some crashes reported on some Windows systems, 
things that were hard to pin down.  I complained to Scott Raney as my 
customers complained to me:  "Scott, if it was a system issue why is it 
that my MetaCard apps are the only ones crashing?"

He pointed me to the Read Me included with MetaCard at the time, which 
included this:

   MetaCard is very good at exposing bugs in the drivers for graphics
   cards.  These problems seem to be most frequent in Windows 98, but
   can plague any Windows system.

It goes on to stress the importance of having the latest drivers 
installed before submitting reports to him.

My first reaction was, "What an arrogant jerk."  After all, if MC is the 
only thing crashing, what the heck could the video driver have to do 
with it?

But he was unwilling to tolerate my tirades until I first tried his 
advice.  So I did.  And on more than 90% of the systems on which errors 
had been reported, once I updated the drivers the problems went away.

I don't know why the engine is so "very good at exposing bugs in the 
drivers for graphics cards", but after that experience I have to agree.

Maybe it's because most of the other apps my customers were running 
(Outlook, Word, Excel, Internet Explorer) had one thing in common:  they 
were all made by Microsoft, and the company has been noted for years to 
use very different APIs for their own stuff than the ones they publish 
for third parties to use.

That's just a guess, I don't really know why.  I just know that what 
Raney asserted was proven true far more often than not, even though it 
defies what seems like common logic.

So FWIW, it may be helpful to at least make sure the graphics drivers 
are current. Your experiences are so unusual and happen across such a 
wide variety of systems that I doubt that's the cause, but at least 
it'll help narrow down the solution space a little more.


> But other times the problems manifest themselves with very simple stacks
> like the one I posted in the Quality Control Centre today (
> http://quality.runrev.com/qacenter/show_bug.cgi?id=7307). Chiildishly simple
> stuff, yet Rev crashes.  Luckily I'd saved that simple stack at one point so
> it could be included with the crash report (but I'd only saved it half-way
> through because I'd already lost it once when Rev hung and I had to force
> kill it).  Feel free to look at that stack.  The problem is, you're unlikely
> to get a crash.  As soon as I restarted Rev I opened the stack, moved the
> field and there was no crash.

I downloaded it and tried it here:  no crash. (MBP 2.16GHz, 2GB RAM, Rev 
3.0build 750, MC IDE)


> None of it is predictable.  The only thing I found that was predictably
> presenting problems was running Bjoernke's Chat Rev on Linux.  Obviously
> that stack works fine for the chat rev users, whatever platforms they are
> running it on (although Bjoernke did note problems with the Chat Rev server
> on Linux -  it would take days for those problems to manifest themselves for
> him, whereas on Linux I could get problems within seconds).

I've used ChatRev often under MC IDE, and a few times in Rev, on OS X 
10.4 through 10.4.11, with engine versions 2.6.2 through 3.0 - no crashes.

I have not run it under Linux, however.  Anyone else here try that?


> I will run without revNavigator for a week and see what happens. 

Worth a try.  I doubt there's anything too subversive in revNavigator 
itself, but perhaps some interaction between what it's doing and what 
the engine or IDE is doing may be the root cause.

Maybe we'll be able to write, "revNavigator is very good at exposing 
bugs in the Rev engine." :)


> The bottom line is this: the engine should not be crashing.

Agreed.

I was at a Microsoft VB seminar many years ago, where the MS rep said 
something very empowering:  "You might see error dialogs, you might have 
unexpected behaviors, but if you ever see a crash that's our fault. 
Ideally, no matter what you do, even if you get sloppy, you should not 
be able to crash VB or the system.  It's our job to make that 
degradation graceful."

That's a lofty goal, and indeed I've never seen any tool, certainly not 
even Microsoft tools, that meets that standard.  But it's worth aiming for.


 > How can anyone trust that they can deliver an application based on
 > that engine, if during development it is randomly crashing?
> If it's not crashing for you, then maybe you feel you can trust it.
 > I certainly don't feel I can trust it.

I can understand your perspective, and if my experience was like yours 
I'd probably be using something else.  But I fall into the latter group: 
Rev crashes very rarely here, and the number of crashes reported by all 
of my customers across all the apps we deliver is very close to zero.

Hopefully once we diagnose the root cause of your anomalous experience 
and correct it, your experience will become as enjoyable.

-- 
  Richard Gaskin
  Fourth World Media Corporation
  ___________________________________________________________
  Ambassador at FourthWorld.com       http://www.FourthWorld.com



More information about the use-livecode mailing list