.9720 floats still truncated when p.linking to ints/choices?

[reezekeys]

That's what my initial testing has shown.

[reezekeys]

On a different computer now (PowerBook G4, 10.5. and things are working fine. Gotta do some more troubleshooting it seems...

That's OSX 10.5.8 above

[dts350z]

Why don't you post a layout or group that demonstrates the issue?

[reezekeys]

Why don't you post a layout or group that demonstrates the issue?

I did post one, down a few threads (".9719 - midi note closer broken?") but I just tried to download it and I get "the selected attachment does not exist anymore." Wasn't deleted by me though.

Right now I am not able to get to the computer that had the issue so I want to wait until I can, then double-check that it's not an error on my part. What I did was pretty simple: put a variable and an int variable on the layout, and used the variable's value to modulate the value of the int variable. As you work the slider for the variable you can see when the int variable changes value.

As I said, my older G4 Powerbook seems to be OK but I do remember seeing the issue with .9720 and my MacBook Pro. If I confirm that it's not my error I will definitely upload a layout for others to check things out.

[dts350z]

By "choices" do you mean linking to "preset number"?

I do see a difference in linking to an int variable's value and linking to it's parameter #

variable (0,1) --> Absolute PM (0,1) --link--> Int value (0,2)

Int changes values when variable is .333333 and .666666

variable (0,1) --> Absolute PM (0,1) --link--> Int preset number (3 presets)

int changes presets when variable is 0.25 and 0.75

but neither of those seem like truncation.

I also tried linking the variable value directly to the int (no PM). Exact same behavior as mentioned above. Are you using a relative PM?

This is on .9720 with windows 7. I can test other platforms (including intel and PPC mac) if you are getting different results than the above.

FYI I use the GUI type "choice" with preset numbers because then the preset names show up as the "choices" in the dropdown.

[reezekeys]

First of all, Merry Christmas/Happy Holidays

I had a little quiet time this morning so I spent it sussing out what is going on with with this issue.

By "choices" do you mean linking to "preset number"?

Yes, and that's what it looks like this is about. Scaling the floats to int variable's value, and scaling them to preset numbers are handled differently, it seems. Whether this issue applies to other choices, I don't know.

Back in March 2010 Seb posted about hidden options in .97, I'll just quote the relevant one here rather than link to the thread:

Presets : added possibility to change scaling
(to have the more natural integer scaling as seen with Variable Int)

OS X:
open a Terminal window:
(to enable)
defaults write com.plogue.bidule Parameter.presetUsesIntegerScaling -int 1
(to disable)
defaults write com.plogue.bidule Parameter.presetUsesIntegerScaling -int 0

Windows :
under HKEY_CURRENT_USER\Software\PlogueBidule
add a DWORD value named Parameter.presetUsesIntegerScaling
set to 1 to have presets behave with integer scaling
set to 0 for default behaviour

*****************************************************

I do see a difference in linking to an int variable's value and linking to it's parameter #

variable (0,1) --> Absolute PM (0,1) --link--> Int value (0,2)

Int changes values when variable is .333333 and .666666

That's exactly what I see.

variable (0,1) --> Absolute PM (0,1) --link--> Int preset number (3 presets)

int changes presets when variable is 0.25 and 0.75

Here's where it's different for me. With Parameter.presetUsesIntegerScaling -int 0, my presets change at 0.49999999 and .99999999. With Parameter.presetUsesIntegerScaling -int 1, the presets change at .33333333 and .66666666

The conclusion, for me, is that .9720 is not returning to the original behavior with scaling the floats to preset choices. That's because, with .9716, Parameter.presetUsesIntegerScaling -int 0, my presets change at 0.25 and .75 like yours. Not so with either pref setting on .9720. This explains my layout breaking.

I also tried linking the variable value directly to the int (no PM). Exact same behavior as mentioned above. Are you using a relative PM?

I get my same results with direct linking. I am recreating your setup, it is definitely an absolute PM.

This is on .9720 with windows 7. I can test other platforms (including intel and PPC mac) if you are getting different results than the above.

.9720 on my MacBook Pro with OSX 10.6.8.

I'm sure this will be sorted out. I'm ready to update my layouts if need be but let's see what happens...

[seb@plogue]

The conclusion, for me, is that .9720 is not returning to the original behavior with scaling the floats to preset choices.

I definitely changed all the code related to parameter value setting and the use of round/trunc to use the old code unless a preference which is off by default has been turned on. Not sure how I could have screwed that one up, I'll check after the new year.

[reezekeys]

Here's a quick and very small video file showing the issue. Can't embed the video, maybe imageshack lets you do that with jpegs only.

I change presets 1-3 on the Int Variable using the Variable above it which goes from 0 to 2. Watch when the presets change:

http://img828.imageshack.us/img828/136/8ke.mp4

Again: I could be wrong, but this looks like truncation, what do you think?

.9720 x86-32 on a MacBook Pro OSX 10.6.8
write com.plogue.bidule Parameter.presetUsesIntegerScaling -int 0

Thanks for looking at this, enjoy the new year and take a break... I am fine with .9716 for now!

[seb@plogue]

That's not what I get here... really strange. Are you linking the value directly to the preset number or do you go through a parameter modulator?

[reezekeys]

That's not what I get here... really strange. Are you linking the value directly to the preset number or do you go through a parameter modulator?

I was linking directly but just re-did the tests using a parm modulator. Same results for me, with a direct 1:1 tracking (variable 0 to 2; parm mod 0 to 2, variable value 0 to 2 or preset 1 to 3).

Here's a few more short videos showing this. I am definitely on .9720:

Modulating the presets, they change at 1 and 2:
http://img713.imageshack.us/img713/5681/2ev.mp4

Modulating the value, changes at around .6666 and 1.3333:
http://img859.imageshack.us/img859/491/14684097.mp4

You said the rounding was processor-specific. I have a 2007 Santa Rosa MacBook Pro, it's an Intel Core 2 Duo T7500. But, .9716 was OK so I'm at a loss. Would the log file give you any clues?

I'm attaching the two layouts I used (of course they're exactly the same except for the mod destination).

[seb@plogue]

You said the rounding was processor-specific. I have a 2007 Santa Rosa MacBook Pro, it's an Intel Core 2 Duo T7500. But, .9716 was OK so I'm at a loss. Would the log file give you any clues?

The CPU rounding mode can be changed by anything else but the default is normally the behaviour as in 9716. Yes I would be interested in seeing the log file.

[reezekeys]

Here is the log file.

[seb@plogue]

OK, it looks like what I packaged as .9720 is not what I expected...

Can you check if this is back to the old behaviour?

http://www.plogue.com/seb/PlogueBidule.zip

[reezekeys]

Yes! Looks like things are back to where they were. Presets change at .5 and 1.5 with defaults write com.plogue.bidule Parameter.presetUsesIntegerScaling -int 0. They change at around .6 and 1.3 with defaults write com.plogue.bidule Parameter.presetUsesIntegerScaling -int 1.

I'm glad this is sorted out! Is this the official .9720 version now, will you update the "latest version" page with it?

[seb@plogue]

It will be up as 9721 when it goes there, I don't want two different 9720 versions...