RE: Re: quit files

Hi Paul 

 I do acknowlege your error report. It is a known error, showing itself in a new light. To be exact, it is one of two major issues that prevented me since April from releasing the SDL Hyperion binaries to the TK4- community (you might have seen that the binaries you are using now are dated March 7th).
 

Because I was never able to reproduce it under TK4- Hercules, I saw this error up to now in the light of something that was newly introduced with using SDL Hyperion. So your weird "pull the plug" usecase has at least one positive aspect: It proves that the error was already present in TK4- Hercules, but is simply occuring with such a low probability that under "normal" usage it never happens. 

 So, what now:
 

 As we all know and have got demonstrated once again just now, race conditions depend on probabilities for the occurance of certain scenarios. Each of the scenarios is in itself deterministic, but there are side conditions that take influence on which scenario is occuring at a specific point in time. As long as those side conditions are not identified, the whole program seems to show arbitrary behavior.
 

 I was trying to shoot this bug down since April (with pauses due to sheer desperation). One of my major problems during analysis was, that reproducability was so weak, that, while having a relatively clear idea where it sits, I never managed to nail it down. Your weird usecase should change this, as it is now occuring relatively often, almost reliably ;-).
 

 Nonetheless, before opening the chase again, I must make sure being really chasing the _same_ bug (as opposed to a similar but distinct one hiding in the same area). In May, I made new binaries containing a kludge with which I was hoping to come closer to the bug. These binaries don't fix anything (because to fix something it needs to get nailed down first) but they change the probability of the occurance of the bug.
 

 So I'd appreciate if you could do the following test for me: Try to find out, with a minimal statistic signifance, how often the bug occurs under
 

 (a) The original TK4- Hercules
 (b) The SDL Hyperion binaries you are using now
 (c) The binaries with the kludge
 

 So, for example, fire up TK4-, then do your weird "quit", and repeat it until the problem has occured 10 times. Then report the number of runs you needed to reach this. Do this for all three versions. Repeat this cycle a few times to see if the ranking is stable. If it is not stable, nothing helps but acquiring more samples, i.e. do it until 20, 30, and so on. Not reaching a stable ranking can of course also be a result. I know that's tedious, sorry about that!
 

 I on purpose don't say what outcome I expect, to avoid biases.
 

 To install the "binaries with the kludge" for (c) download https://polybox.ethz.ch/index.php/s/OrtQJFQaScNJN2c/download https://polybox.ethz.ch/index.php/s/OrtQJFQaScNJN2c/download and unzip it into the tk4- folder that has your current SDL Hyperion binaries activated.
 

 Thanks in advance for reporting back.
 

 Cheers
 Jürgen

---In [email protected], <kerravon86@...> wrote :

 ---In [email protected] mailto:[email protected], <kerravon86@...> wrote :
 
 > 1. Unzip standard TK4-8
 > 2. Unzip below download
 > 3. Make this change:
 
 Sorry, there was also:
 
 2.5 Run activate_SDL
 
 BFN. Paul.