Possible Value Curve Conversion Problem in 2017.27

[robskimeister]

I recently upgraded from 2017.23 to 2017.27 and I'm seeing some issues with some of my effects that use value curves.  Has anyone else noticed any strange behavior with value curves after converting pre 2017.24 sequences to the newer versions?  You may want to take a closer look at your value curves to be sure.

I've included some screenshots of one of my "problem" effects showing the pre-2017.24 value curves, and the converted 2017.27 values.  They don't look to be equivalent.

Also, the effects visibly appear totally DIFFERENT in 2017.27 than earlier versions.  I'll post a video later if needed to show the difference.

I'm afraid there may be a glitch or bug in the conversion calculations.  If so, that means people who use value curves may have corrupt sequences if they're using 2017.24 or later and may want to hold off on sequencing in the newer versions.

I can file a ticket for this issue on Github if needed, but I figured that I'd see if any others have noticed this before doing so, in case I'm doing something wrong.

Please let me know if any files or other information are needed.

Thanks again to all the developers who make xLights possible!   

[keithsw1111]

Can you package your original .23 sequence for me ideally just with the effect that demonstrates the problem.

[robskimeister]

Can you package your original .23 sequence for me ideally just with the effect that demonstrates the problem.

Thanks Keith!

Here are a few of the problematic effects in a packaged sequence file.

Just an FYI:  I've had issues with other effects as well, but I didn't think there was an issue at the time and didn't document which ones they were.  I figured I probably goofed up when I created them.  But I knew the effects I posted were wrong when I saw them since they were all so messed up.  So the issue is likely in other effects as well.

I was really hoping others would be having similar issues that would help determine the cause.  Surely I'm not the only one using value curves.

Thanks again for looking into this! 

[keithsw1111]

For clarity these are the .23 versions before conversion ... right

[robskimeister]

For clarity these are the .23 versions before conversion ... right

Yes, before conversion loaded from my backups.  These were packaged in 2017.22.

[keithsw1111]

As best i can tell it only impacts curves where the value supports a decimal point. Fix will be in .28

[robskimeister]

As best i can tell it only impacts curves where the value supports a decimal point. Fix will be in .28

Cool!  Hopefully that shouldn't affect too many of them.  Thank you!

[robskimeister]

Hey Keith,

Is there any easy way to search within the sequence XML files (my backups before conversion) and find only the effects that would have converted incorrectly?  What terms would I need to search for?

I'm asking because I've converted 2 songs and performed a considerable amount of sequencing in each of them since conversion.  So reconverting them and losing all that work really isn't an option. 

I'm hoping I can convert the backups after the issue is corrected, and copy/paste the corrected effects into my primary sequences.

I hope that made sense.

If not I'll just watch the preview & see if I spot anything odd.

Thanks again for all the hard work!

[keithsw1111]

The good and bad news I don't think there is an issue unless you click on the effect with the curve and then it should be obvious

This is also bad news because I now realise that will be a problem for me. Not you. I will need to code a forced upgrade soon or I will have problems in the future.

This only affects a handful of values where the value can be fractional.

There is no way to search the xml for these.

[robskimeister]

The good and bad news I don't think there is an issue unless you click on the effect with the curve and then it should be obvious

This is also bad news because I now realise that will be a problem for me. Not you. I will need to code a forced upgrade soon or I will have problems in the future.

This only affects a handful of values where the value can be fractional.

There is no way to search the xml for these.

OK, I appreciate the info.  Hopefully you'll be able to get the code fixed without too much trouble.

Thanks again for looking into this issue.