[1.8.0 RC5] Stability Problems Solved

[Juan Conejero]

Hi everybody,

In first place, thank you so much for helping me with this issue. It seems that I have managed to identify and fix this insane "crashing closing windows" problem.

Two repeatable instances have been crucial to find the solution:

* Andrés del Pozo found a sequence of operations that invariably leads to a BEX64 crash on PI 1.8.0 RC5 for Windows:

1. Open any image
2. Open the HistogramTransformation tool
3. Activate the Real Time Preview window from the HT tool
4. Maximize the RTP Window
5. Modify any HT parameter
6. Apply HT
7. Close RTP
8. Maximize the image where HT has been applied in step 6
9. Close the image -> BEX64

For the problem to be 100% repeatable, the nine steps are necessary. Andrés uploaded a video that demonstrates the problem.

* Martin Magnan discovered the following sequence, which invariably leads to an application crash on Mac OS X with PI 1.8.0 RC5:

1. Open an image
2. Duplicate the image (or, alternatively, open a second image)
3. Open the LinearFit tool
4. Select the image from step 1 as reference image on the LF tool
5. Apply LF to the image from step 2
6. Close the image where LF has been applied in step 5 -> Application Crash

Martin also recorded a video showing the problem. This sequence shows that the LinearFit tool simply cannot be used in RC5 (probably also on RC4 and RC3). We should have discovered this before.

Finally, I have been able to reproduce the bug on both Windows and Mac OS X, on all of our development machines, following specific steps that consistently lead to the same crashes that you are experiencing. And happily, I have managed to find where the problem is, and fix it.

After the fix, not only the above sequences can be executed normally, but all the recent problems on Windows have disappeared automagically: The DarkStructureEnhance, DeconvolutionPreview and RepairedHSVSeparation scripts, as they have been distributed with RC5, work now flawlessly on Windows. There are no more BEX64 errors or access violations, or at least I have been unable to reproduce them in all cases where they were consistently reproducible before.

I'll publish an RC6 version during this week, which should be fully stable on all platforms. Thank you all for your patience and support.

[Alejandro Tombolini]

What an excellent news Congrats!!!

[georg.viehoever]

This is really good news! Thanks to all those who helped to track down these beasts. And i will email  a bottle of wine to Juan for his hard work!
Georg

[Josh Lake]

What a relief! Thanks for the hard work. I'm sorry I didn't get the chance to send in my own crash reports, but I had had a few.

[f11]

Many software developers would have just thrown up their hands, blamed the OS, and walked away when faced with a problem like this.

While PI is an unbelievably powerful tool which you and your team can be very proud of, its your stubborn pursuit of complex bugs like this that sets PI aside from its lesser wanna-be's.  Your dedication to your craft is what we as consumers pay for when we purchase PI.

Thanks for your efforts, and BIG THANKS for fixing the crashes!
Rod

[federico]

Juan, do you think that the new release will solve similar problems under Linux? As I said, I've had a lot of crashes after window closing in Ubuntu 12.10.

[Jan]

Not only the great software, but also amazing support, thank you Juan! I do not regret a penny spent on Pixinsight, wish other products I use had such a great team behind.

[Komet]

Excellent news but I would like to point out that I did report the steps to reproduce this bug before any video and before which I was told it was probably the OS - which,  to be honest, I believed at the time. It was only when I was able to reproduce it was I able to point the finger at PI.

No matter .... excellent news Juan, Well done and thanks.