Bug reports

[Robbert van de Kruijs]

Let me first start by thanking David for his amazing efforts with the new and improved(!) IMD!

 

David, since you suggested to have feedback in this group, I wanted to start a topic on "bugs", I hope that's okay?

I suggest that all posts to this topic include clear information on IMD version/installation type(location)/...

 

Alright, since it's hard to know what is a bug and what is a feature, please forgive me if I address not a bug, but a feature... :)

 

---

IMD5.01 installed on 5y old Dell PC, WinXP, location "c:\program files\XOP2.3\..." (new install of XOP2.3)

---

 

- During a fit, I got a "Window: Unable to create pixmap" after a day of genetic algorithm.

Fit was paused with this screen until I clicked the error and then it just continued...

Did anyone encounter similar things?

 

 

Actually, no real major issues, just minor things so far, I'll report more if and when I encouter them, happy to provide more details when requested.

 

Cheers,

Robbert van de Kruijs

[David Windt]

Hi Robbert,

Thanks very much for all your feedback, in this and the other messages you just posed - much appreciated!  I'll respond to each message separately.

This is definitely a good place to report bugs, so no problem starting a 'bugs' thread. (Though it's not necessary to use this thread to report ALL bugs - that might quickly get messy and confusing.) Anyway, the issue you report here is certainly not a feature! Though I'm not entirely sure it's a bug either, in the sense of being a coding error in IMD. It might be, but I'm not sure I can tell for sure from what you've reported. I don't recall seeing the 'Unable to create pixmap' message during a fit as you've apparently encountered. But a message like that in IDL can mean that there was a problem with IDL running out of memory, or perhaps IDL ran out of pixmap indices for some reason. In any case, if it is a bug, I would certainly like to fix it.  I don't suppose this problem is repeatable for you? If it is, then please send me the .imd file that will cause the problem. If it's not repeatable it will be trickier to fix, but please do let me know if it keeps happening and we can try to track down the problem.

David

[Robbert van de Kruijs]

Hi, David,

 

Yeah, could be memory related, and since they were >1 day going in the weekend before it crashed, I'm afraid I did not reproduce the fits on Monday...

(and of course did not save the .IMD file before fitting, so no chane to reproduce...)

I'll be sure to automaticall save stuff and send you the files if things can be reproduced, but so far this was the only crash, and it may somply be my old PC, so you gave me a good reason to get a new one :)

 

Robbert

[Patrick Kearney]

David, still exploring all the new features, having too much fun for my boss.  I did run in to an annoyance.  I was setting up a simulation and was making a coupled parameter so two layers would have the same total thickness as one was varied.  Basically I had a coupled parameter like Layer2=100-layer1  (thicknesses) and layer1 was an independent variable.  In my haste I got a number wrong and ended up with a negative thickness and got an error message.  No problem so far.  But I kept getting the error message after I fixed my mistake.  I think I even had to exit IDL and restart to make it work again.  So, I got it to work, but there seems to be some problem with the initialization of the error check. Couldn't save my structure either.  So, I lost a few minutes work. 

 

Thanks again for all the hard work, I owe you a beer (or name your poison)

Patrick Kearney

 

[David Windt]

Hi Patrick,

Looks like you found a bug in IMD - I'll aim to fix it for the next release. In the meantime, you can avoid the negative thickness error by deleting the coupled parameter that's causing the problem.

Best regards,

David

[Dmitry Kuznetsov]

Hi, David.

Today I've catched a fact of strange behaviour of IMD, that is perfectly reproducible on my PC (for info, OS Windows x64), but within one file.
https://drive.google.com/file/d/0B6T92xyewYcpUW5KNUxJRkxzT28/edit?usp=sharing
(I've created exactly the same MLS in IMD from scratch - no more error)
IMD keeps initial values of fitting parameters, writes that chi square is infinite and do not change any parameters at all,
generating the same generations in time...
Below is the link to download the buggy .imd  file:
https://drive.google.com/file/d/0B6T92xyewYcpOWZvTFVRdWNrdm8/edit?usp=sharing


Best regards, Dmitry S. Kuznetsov.

[David Windt]

Hi Dmitry,

It looks like you have a bad measured data point - one of the points (y_m[136]) has a value of 0. Delete that data point and the fitting should work normally. I'll adjust the code to make sure that such bad data points don't cause this problem in the next release.

Best regards,

David

[Dmitry Kuznetsov]

David,

I agree.
I've checked the measured data, and there are some points of R values like -0.000063,  -0.000022, -0.000108...
IMD sets them to zero, and if the exp. data (lambda) range is chosen so that there are no such points, error does not occur.

Thank you for the time spent!

Best wishes, Dmitry  S. Kuznetsov.

[Dmitry Kuznetsov]

Hello, David!

Today I catched an error I've never seen before in IMD 5.02.
After I've added layer thickness errors to Independent Variables, made calculations,
then closed graphs, deleted that parameter, and then tried to calculate reflectance for the MLS
without this parameter. The latter brings me to the error
https://drive.google.com/file/d/0B6T92xyewYcpUllLRTU4Sy1zOWc/edit?usp=sharing

I'm not able to reproduce the error intentionally,  but few times it popped with that one particular file
https://drive.google.com/file/d/0B6T92xyewYcpUmtfMVZBZnd1U00/edit?usp=sharing

Wishing all the best,
Dmitry Kuznetsov.

[David Windt]

Hi Dmitry,

Did you change materials and/or wavelengths just before you saw the error? There's a known bug that might explain the error you see, which I will correct as soon as I have a chance...

David

[Dmitry Kuznetsov]

Cannot tell now for sure if I did change or not, but now, even if I change wavelength and/or materials, no bug shows itself.