Explicitly handle VST2 callbacks with nullptrs

This is probably not needed, but we'll do it just in case another plugin
does something similar.

CHANGELOG.md

@@ -13,6 +13,12 @@ Versioning](https://semver.org/spec/v2.0.0.html).
 - When building the package from source, the targetted Wine version gets printed
   at configure-time. This can make it a bit easier to diagnose Wine-related
   compilation issues.
+- Yabridge automatically handles most common VST2 functions by simply inspecting
+  the argument types. This works practically everywhere, but Plugsound Free by
+  UVI would pass unreadable function arguments to functions that are not
+  supposed to have any arguments, causing yabridge to crash. To prevent similar
+  situations from happening in the future, yabridge now specifically handles
+  most common VST2 functions that don't have a data argument.
 
 ### Fixed
 
@@ -22,7 +28,7 @@ Versioning](https://semver.org/spec/v2.0.0.html).
 - Prevent _Native Instruments' FM7_ from crashing when processing MIDI. As a
   fix, MIDI events are now deallocated later then when they normally would have
   to be.
-- Prevent _Plugsound Free_ by UVI from crashing during initialization.
+- Fixed _UVI Plugsound Free_ crashing during initialization.
 
 ## [3.1.0] - 2021-04-15
 

src/wine-host/bridges/vst2.cpp

@@ -490,13 +490,33 @@ class HostCallbackDataConverter : DefaultDataConverter {
             case audioMasterGetProductString:
                 return WantsString{};
                 break;
-            // HACK: DefaultDataConverter::read() should be able to handle all
+            // NOTE: DefaultDataConverter::read() should be able to handle all
             //       of these 'simple' opcodes, but Plugsound Free by UVI passes
-            //       random garbage for their data argument, which we would then
-            //       try to read as a string resulting in a memory error.
+            //       random garbage for their data argument. Because of that
+            //       `audioMasterWantMidi()` will segfault because when we'll
+            //       try to read that data as a string we'll start reading
+            //       unallocated memory. Even though no other plugins seem to do
+            //       this< we'll list all of these data-less opcodes just to be
+            //       sure. We're leaving out a few opcodes here, because I have
+            //       no clue whether some of the more obscure ones are supposed
+            //       to have an data argument or not.
+            case audioMasterAutomate:
+            case audioMasterVersion:
+            case audioMasterCurrentId:
+            case audioMasterIdle:
+            case audioMasterWantMidi:
+            case audioMasterSizeWindow:
             case audioMasterGetSampleRate:
             case audioMasterGetBlockSize:
-            case audioMasterWantMidi:
+            case audioMasterGetInputLatency:
+            case audioMasterGetOutputLatency:
+            case audioMasterGetCurrentProcessLevel:
+            case audioMasterGetAutomationState:
+            case audioMasterGetVendorVersion:
+            case audioMasterGetLanguage:
+            case audioMasterUpdateDisplay:
+            case audioMasterBeginEdit:
+            case audioMasterEndEdit:
                 return nullptr;
                 break;
             default: