User Tools

Site Tools


extending_the_effects_engine_report_w11

Hello again,

Last week I told you that I added the handling of Lilv library inside features.py. There I set the minimum version of lilv to be 0.14. Daniel pointed me out that Ubuntu 12.04 repositories still feature the old lilv 0.5 version. I looked for that version of lilv and looked inside lilv.h. All the functions I used for adding LV2 support to Mixxx are there, so I changed the minimum required version of lilv to 0.5 inside features.py.

As a side task I implemented a 10 band Master Graphic Equalizer. The work can be broken down on two parts:

  • Create a new effect which features 10 bandpass static filters. Their frequency is set up when they are created. Biquad Bandpass filters generated with fidlib are used for this effect. This new type of filters is based on Daniel's IIR filters refactoring[1].
  • Create sliders in preferences which control the newly created effect's parameters.
    • These sliders' sliderMoved signal is connected to a slot which is updating the effect parameters directly, by writing requests to the Effect Engine.
    • The naive way of doing this is having a different slot for each slider:
      • slider1' sliderMoved signal connected to slotSlider1
      • slider2' sliderMoved signal connected to slotSlider2
      • slider10' sliderMoved signal connected to slotSlider10
    • However, a lot of code is duplicated, because there are only 2 values which differ from one slot to another. I had to find a way to have access to a couple of values inside the slot responsible for updating the state of a parameter:
      • the index of the slider which was modified –> for knowing which effect parameter to update
      • the new slider position –> for obtaining the value which needs to be set for the effect parameter (this is passed by the sliderMoved(int) signal)
    • I tried to use QSignalMapper and map each signal with an integer representing the slider's index. Unfortunately, QSignalMapper works only on parameterless signals.
    • The best solution was making use of QObject's dynamic properties. I used setProperty method to store the slider's index. All I had to do inside the slot now was to use the sender() method to obtain the slider which emitted the signal and get retrieve index which was previously stored as a dynamic property.

Here is a screenshot with the current state of the Master EQ:

I have some good news about the kn0ck0out LV2 plugin. As you know from my ninth report, it was not working as it should have, it was only playing silence. So I contacted its maintainer, Jeremy Salwen. He was kind enough to take a look on the code and found out that the output of an integer division was used instead of a floating point one. Here[2] you can check out the commit. I'm flattered he mentioned my name in the commit message.

I continued my work on LV2 support by adding enumeration and button parameters. They are based on a multi state button (two in case of a toggle parameter, more than two in case of an enumeration parameter). The main thing this type of parameters needed was for their underlying value to go back to the minimum if the maximum was exceeded. This[3] commit implements that behaviour. Working on this I got a “strange” C++ error which cause was using a getter method on a const object. All I had to do was make the getter const, because only const methods can be called on const objects. This makes sense because otherwise a non const method might attempt to modify the const object.

After testing some LV2 plugins and their enumeration parameters, I was not pleased to realize Mixxx crashes chaotically when changing effects. What made this bug hard was that I got different errors each time I ran Mixxx and using debug mode did not help me very much. I found out that the bug disappears if I remove the code responsible for updating lv2 parameters: params[i] = m_parameters[i]→value().toFloat(). params is a dynamically allocated float array which is connected to lv2 plugin instance's ports. The problem was that I allocated memory only for manifest.parameters.size() and forgot to take into account manifest.buttonParametrs.size(). C++ does not perform boundary checking. Thus, sometimes the code even worked, maybe because the memory area right next to params was not accessed by someone else. This bug taught me to be more careful when allocating resources in C++.

Cheers,
Nicu Badescu

[1] - https://github.com/mixxxdj/mixxx/pull/294
[2] - https://github.com/jeremysalwen/kn0ck0ut-LV2/commit/7ff733d90e6c51649203ab799977a8be616f08fb
[3] - https://github.com/badescunicu/mixxx/commit/0a327759d4bc91f25f60059e29f222436dc9b719

Back to the main page!

Translations of this page:
extending_the_effects_engine_report_w11.txt · Last modified: 2014/08/03 18:25 by badescunicu