LMMS
LB302 - no sound
System Information
Debian Stable, wine32 on Debian Stable, Windows10 64-bit
LMMS Version(s)
lmms-1.3.0-alpha.1.667+g1c865843f
Most Recent Working Version
For LB302: point just after 2024-03-28 19:13:32 Fix glitch in SlicerT #7174
Bug Summary
On start working version(1.102), but the 2-nd part - compiled master (the same with win64 version on Windows 10) https://github.com/user-attachments/assets/5067ce77-760a-4a21-acc7-14ba4647cb9b LB302 is OK if use win32 version installed using wine (but SID notes with clicks)
Expected Behaviour
Sound without clicks. But: No sound (LB302), (clicks on every note start (SID))
Steps To Reproduce
Open (attached) project: lay with LB302 and SID
Logs
Click to expand
Screenshots / Minimum Reproducible Project
Please search the issue tracker for existing bug reports before submitting your own.
- [X] I have searched all existing issues and confirmed that this is not a duplicate.
Reverting "Switch to libsamplerate's callback API in Sample" #7361 not fix bug - this bug is NOT connected with #7361
My future plans (Lento Tranquillo , but in parallel): (1) find point, there this regression happens (I'm not familiar with git - this is good practice for me); (2) some how replace LB302 (SID) code in master using pre-release code.
Just compiled at point just after 2024-03-28 19:13:32 Fix glitch in SlicerT (#7174) : LB302 working, but SID is with clicks. P.S. For LB302: pre-release code no more actual.
It seems is "Mate in 2":
(in folder plugins/Lb302/)
if file Lb302.h (~217 line):
should change
int release_frame;
to
f_cnt_t release_frame;
In file Lb302.cpp (~756 line):
should change
release_frame = std::max(release_frame, static_cast<int>(_n->framesLeft()) + static_cast<int>(_n->offset()));
to
release_frame = std::max(release_frame, _n->framesLeft() + _n->offset());
(as it was before, except std::max)
P.S.:
But I wander: why release_frame has so big value:
Before: ~1073741823 (int32_t)
Now: ~9223372036854775807(size_t : unsigned long)
It is something wrong to give NotePlayHandle object so big working parameter values ...
For SID: "conflict wall" at point after "Fix uninitialized variable "pCurPreset" (#6639) (#6723)" (2023-06-05 06:44:48) Can not compile : new fetched code has conflicts with this revision. P.S. In new master bug is not "magically disappeared".
Now I compare Master with PreRelase : 1-st iteration do not give a result ... (1) My opinion - old code more readable, compact and elegant, than new "cool C++ style", but it should work in the same way as old code. (There are some points, but if bug is there - sound effect should be different.) (2) On every note SID send some kind of pattern - and this make click. This is old chip emulator and I do not like this on my flute like setting, but may be this is not bug but emulated chip sound? (3) May be bug is in plugins/Sid/resid/ , downloaded pre release zip not contain this ...
I will pause 10 days waiting answers: (1) I gave enough information to fix Lb302 no sound bug. Can someone, who has write access simply commit this 2 line changes to master? (2) Can someone from maintainers/members (best variant - SID authors) confirm, that SID sound with start clicks now is a bug? (and SID should sound as in pre release) (Edited:)P.S. With Jack SID now introduce:
JackEngine::XRun: client = lmms was not finished, state = Triggered
JackAudioDriver::ProcessGraphAsyncMaster: Process error
port created: LMMS:midi/capture_22
ALSA-sequencer: unhandled input event 66
...
But no problems in pre release ... So SID is buggy and it is serious regression ...
So 10 days over and I have no SID sound bug confirmation: assuming SID should produce clicks on every note start.
SID XRun bug should be another issuer opened, and this XRun can be reproduced using SDL audio back-end as project sound defect (really - pause) on 1-st SID note played.
I will make new issue soon, than edit this one.
Now my plan is to try to find way to fix "why release_frame has so big value".
This big value now are negative values castled to unsigned long (size_t).
Mate in 2 fix "chess blunder" in :
Define fpp_tandf_cnt_tto be ofsize_t (#7363)
Such change introduce a lot of compiler errors and warnings, at first time (some of) were simply castled. (Refactoring PR) Author forgot about this one line. Also Mate in 2 make release_frame "type consistent" : LB302 implementation author ignore lmms_basics.h type using C/C++ base type - "base type" for f_cnt_t before Define fpp_tandf_cnt_tto be ofsize_t (#7363) .
But it seems, that f_cnt_t:int32_t -> size_t introduce some hidden problems ... Joke is that in 90% CPU treat signed and unsigned types in the same way - so code is working even if underflow happens by negative values used before.
But there are 10% ... One of the such signed and unsigned difference happens, than type size is expanded.
On 32-bit platforms sizeof(size_t) == sizeof(int32_t), but this is not true on 64-bit platforms.
May be this is a answer to question: "Why LB302 have sound on MinGW32 compiled version".
It seems that Fix Lb302 silence #7504 should fix this bug, but I have not tested this yet.
P.S. PR is closed, issue not mentioned, so I connect ...
It seems that Fix Lb302 silence #7504 should fix this bug, but I have not tested this yet.
Yes. The fix is already in master. This issue is closed with https://github.com/LMMS/lmms/pull/7504
