AutoSTF crashes, likely LGRBCombination problem

[bitli]

Hello,

I have a case of STF crashing PI when doing an AUTO on a small image. This is systematic with that image (unfortunately even compressed it is larger than 192KB). The image was produced by the LRGB combination, using the RGB as a target, replacing only L, then doing an Apply Global.
When I do an 'drag of new instance' to the target, the resulting image can be STFed normally.

So I suspect that the LRGB combination does not produce a correct image if apply global is used. The work around is obvious, however it should not create an image that crashes PI.

If required I can mail the images (about 1MB).

-- bitli

[bitli]

To be more precise, PI does not crash, it goes to an endless loop and I have to kill it. This is on Windows 7 64 bits with a 4 core cpu.

-- bitli

[Carlos Milovic]

Do you have any pixel that reads "NAN"?

[bitli]

I have put the crashing file as well as the two files used to build it in
http://lugrin.ch/astro/images/CrashAutoStf.zip
for a limited time.

There are values listed as 1.#QNB, what ever that is.

-- bitli

[Juan Conejero]

Hi bitli,

Thank you for uploading the images. They have allowed me to identify the problem perfectly.

You have now a record in the history of PixInsight bug tracking: you have discovered three bugs for the price of one  8)

- LRGBCombination should not allow you to generate an image if you don't select existent (and valid) L, R, G and B source images.

- LRGBCombination is using garbage data (arbitrary data gathered from your computer's memory) to form a RGB image. The core application should detect out-of-range and NAN values and fix them.

- ScreenTransferFunction should not crash in presence of NAN (not-a-number) values (which M$ denotes as "1.#QNB" for some exotic reason).

More work for 1.6.1. Sweet!

[Niall Saunders]

(which M$ denotes as "1.#QNB" for some exotic reason)

Juan,

I have been seeing this on the new (beta) "ReadPSF" module, and was wondering what was the cause of these 'illegal value' representations !!

What are you referring to by <<M$>> - is this Micro$oft ?!!?

Cheers,

[Juan Conejero]

Hi Niall,

What are you referring to by <<M$>> - is this Micro$oft ?!!?

What's making you think that?

[bitli]

When you are at it, can you check if droping a monochrome stf instance on an histogram work? It seems to me that it does not work (it works for color sourcrs...)
-- bitli

[Carlos Milovic]

I have been seeing this on the new (beta) "ReadPSF" module, and was wondering what was the cause of these 'illegal value' representations !!

Oh. This is done by a zero division I need to add a routine to calculate a "internal" threshold if it is set as zero (defined by the sample radius).