He wanted to help so I asked if he wanted to track them down. He agreed and
found more than I probably would have. And he said there were more, but the
rest were questionable and since he works on Windows, wasn't able to actually
test.
I give a few of the ones he did find a go and they do indeed get rid of the
assertions.
(Basically, a 64-bit/32-bit issue, easily resolved with a typecast.)
This relies on three new nyquist scripts to actually do the editing. The peak-snapping code in FrequencyWindow has been extracted into a new class, SpectrumAnalyst, to provide peak-snapping in spectrogram too.
With the addition of the preview button, it became evident
that if an effect were to return text instead of audio, the
processing loop would not stop properly.
And he also found a case where the debug state wasn't reset
after using the debug button. If this was following by the
preview button the next time the effect was used, the debug
window would reappear before the preview.
Steve found that if you preview an effect and, say, cancel it
before it finishes preparing, an apply will not modify all of
the originally selected audio...just up to the preview length.
This is very easy to reproduce using the Paulstretch on a 30
second clip...just click stop on the "Preparing preview"
dialog and the click apply.
We can't go to 3.0.1 yet as there are still build issues on
Linux and OSX. You can get Windows to build, but there's
still some display issues.
These changes should work with wxWidgets 2.8.12 as well, so
we can take our time to get things working properly before
switching over.
He also added a new Nyquist control header that gives each
effect the ability to control the presentation of the Preview
button. This will be used by effects are time based and, therefore,
do not produce the same results in preview as they do when applied.
I expect Steve will be committing changes to the effects that can
take advantage of this shortly.
Testing of this can be accomplished by adding:
;preview enabled
to the "tremelo.ny" effect, just after the ";action" line.
Adding that line will cause the "Preview" button to be shown and
you will then be able to preview, adjust, preview, apply...
This change is believed to be a direct refactoring that does not change functionality. It paves the way for more complex kinds of selection, such as selections involving frequency as well as time. It also reduces risk of left and right edges being swapped in future code using SelectedRegion, as the default is to swap on assignment if needed.
I can't believe two things...
I can't believe I didn't test the textual display! I'm really ashamed about that.
And I can't believe that we've never had anyone report a crash when attempting to
using the textual display with plugins that do not supply textual parameters. Some
examples are the ones from BBE Sound.
I went back to Audacity 2.0 and they would crash there as well. I'd wager they
probably always caused Audacity to crash.
Not much can be done except to fall back to the graphical display.
After reporting the crash to the kn0ck0ut LV2 maintainer, he pointed out that
Audacity wasn't connecting to all required LV2 ports. In this case it was the
latency port. This corrects that oversight.
In addition, since I was fiddling with the latency port, I figured I might as
well go ahead and add the necessary code to compensate for it.
Bugs fixed:
1) When scanning for VSTs on Mac, bogus lines could show up in the list if the .vst
file had more than one Info.plist in the bungle. GRM and TAL plugins exhibited
this issue.
2) When scanning for VSTs on Mac, reuse of a variable caused the plugin to be
accepted even though it was a synth. So it would show up in the Generate menu.
3) I recently disabled the the plugin list dialog once the OK button was clicked
to prevent clicking it again. This was overkill and prevented clicking Cancel
while the scan was running. So, changed it to only disable the OK button.
4) The new Settings dialog would save the settings even if you clicked Cancel.
5) Detecing the overlay window on Mac didn't work in all cases. Did more research
and I believe the revised method is correct. (Window list is z-order sorted)
6) When changing the preset from the top combo, the return value from the
plugin was not check when getting the current preset (program). It assumed
that there was always an active program. This is not at all correct.
Most of the new additions are an attempt to be more compatible with more plugins.
I have not found any plugins that crash Audacity anymore, so I'm not sure how to
really give these changes a good workout. However, Guitar Rig 5 and Reaktor 5
are now quite usable since one of the additions was window resizing.
1) Refreshed the aeffectx.h header from the LMSS project and merged in some
additional changes from Ardour.
2) Changed the method prototypes to reflect the new aeffectx.h header.
3) Plugin scan dialog now has a "Select All" and "Clear All" buttons. (On a
personal note, this was a major boon while testing...way easier to include
a single VST.)
4) Uses a timer now instead of the wxIdleEvent to send effIdle messages to the
plugin. Idle could be all many times more than was necessary.
5) Process level is now reported when requested by the plugin. Don't know how
useful it is, but many plugins request it.
6) Two new events are sent to the dialog from the audioMaster callback to support
window updating and resizing. Figured it was safer this way since the callback
"might" be called from a different thread (as far as I could figure out anyway).
7) Current sample rate is now returned to the plugin when requested.
8) The dialog now resizes when the plugin informs us of a size change to its window.
This was one of the major drawbacks to using Guitar Rig and Reaktor as their
window would resize and pretty much be useless since the dialog didn't adjust to
the new size. Kontakt was another that even includes a draggable resizer in the
bottom right corner which the dialog supports just fine now. (Not that Kontakt
is actually supported though.)
9) The preset bar is now updated when the plugin informs us of "major" changes with
the UpdateDisplay opcode. I noticed that when changing presets within Guitar
Rig, the preset bar didn't change.
Discovered limitation:
Guitar Rig and, it seems, other plugins will only report a maximum of 128 presets (or
rather programs) even if they support more. I'd figure out a way to populate the
combo with all of the presets via some major hackage, but it didn't do any good since
Guitar Rig simply refused to set any program number above 128.
Not much can be done about this...just a heads up.
The buffer delay compensation resolves the issue where an effect
may add a delay to the output. This would cause the output to shift
and the final samples to get chopped off.
The settings (click Settings on any VST dialog) dialog allows the user
to enable/disable the BDC and to set their desired buffer size (within
reason) to help improve performance.
In addition it fixes a bug in the plugin scan list were you could click
OK twice and have two scans going, eventually asserting.
Fix a typo in the name of 'Classic Filters'.
Improve consistency layout as we switch filter types.
Remove a win warning in EQ.
Improve some code readability / change styles (please don't shoot me for that!). OK, possibly increased the inconsistency.
