Consider disabling optimizations for SWH Multiband EQ

[tresf]

Currently, we use a CMake-adapted version of Steve Harris' Makefile and use optimizations -03, but several upstream bug reports/upstream comments suggest this could cause some audio artifacts under certain configurations, notably some 32-bit hosts.

The symptoms are crackling/frying noise. Oddly, not all LADSPA hosts exhibit the problem, quoting:

Bit of a dredge... but I am also getting EXACTLY the same noise/hiss/crackle as you are playing your ogg file.

it is strange - since aplay and mpg123 play fine directly to alsa configured alsaequal or direct LADSPA plug ins. This works for zm1, vynil and mbeq on my ArchlinuxARM system.

However - when i use mbeq under MPD (either via alsaequal or ladspa) - I am getting the noise you are.

This was working fine on my old ArchlinuxARM version - so I'm guessing a downstream package upgrade has broken things... I've tried "forcing" downgrades - everything works (until you really go into the bowels) - e.g. alsa-lib, fftw, gsm, yet the hiss from mbeq is there.

I can get all plugins working via mpg123 and aplay clients. MPD works with vinyl via LADSPA direct, or zm1, but hisses / crackles via LADSPA or alsaequal under MPD (same asound.conf). The funny thing is MPD works fine with alsaequal's 10 band EQ (in caps.so) under MPD.

Upstream bug reports:

• mbeq plugin outputs frying noise when I use 32 bit library swh/ladspa#64 (comment)
• mbeq plugin flat mode volume differs from volume without the plugin swh/ladspa#70 (comment)

lmms/plugins/LadspaEffect/swh/CMakeLists.txt

Line 12 in 4dc04a6

SET(COMPILE_FLAGS "${COMPILE_FLAGS} -O3 -Wall")

There's really no immediate action needed here on LMMS' end, I just found the issue to be popular enough that eventually someone will run into it downstream on our project and sometimes these optimization issues are really hard to track down. 😅