Faust 2.58.11 to test

sletz

OK, I was able to reproduce the problem on a test program.

@Christoph-Hart you'll have to explicitly set the target parameter to "x86_64-apple-darwin" when using the c reateDSPFactoryFromFile/createDSPFactoryFromString
functions.
(see https://github.com/grame-cncm/faust/blob/master-dev/architecture/faust/dsp/llvm-dsp.h#L179

Can you test on your side and report?

Christoph Hart

@sletz Now it compiles but crashes at createDspInstance() / llvm_dsp_factory::getWarningMessages()

Log:

Faust: Resizing buffers: nChannels=0, blockSize=0
Faust: Resizing buffers: nChannels=0, blockSize=0
Faust: Resizing buffers: nChannels=0, blockSize=0
Faust: Resizing buffers: nChannels=0, blockSize=0
Global Faust Path: /Users/christophhart/Development/faust
Global Faust Path: /Users/christophhart/Development/faust
Faust jit compilation successfu::

Assembly stack trace (not expecting to be helpful, but anyways):

    0x1113b99f0 <+0>:  pushq  %rbp
    0x1113b99f1 <+1>:  movq   %rsp, %rbp
    0x1113b99f4 <+4>:  pushq  %rbx
    0x1113b99f5 <+5>:  pushq  %rax
    0x1113b99f6 <+6>:  movq   %rdi, %rbx
->  0x1113b99f9 <+9>:  movq   0x18(%rsi), %rsi
    0x1113b99fd <+13>: movq   (%rsi), %rax
    0x1113b9a00 <+16>: callq  *0x58(%rax)
    0x1113b9a03 <+19>: movq   %rbx, %rax
    0x1113b9a06 <+22>: addq   $0x8, %rsp
    0x1113b9a0a <+26>: popq   %rbx
    0x1113b9a0b <+27>: popq   %rbp
    0x1113b9a0c <+28>: retq   
    0x1113b9a0d <+29>: nopl   (%rax)

sletz

Same issue using the Interp backend, although this getWarningMessages method is not called at all.... No more idea for now. I see libfaust is still called in a separated thread ?

And what happens on a real Intel machine ? (the compiled version is for macOS 10.15 minimum, and I still have a 10.13 here where I could not test).

sletz

I've found the problem: changes in the Faust dsp/dsp_factory API, and since HISE is maintaining it's own version, this make the thing quite fragile.. @Christoph-Hart I'll send you a patch.

Christoph Hart

@sletz Yes I can confirm that the patch you send me works now (I had to add the darwin flag but apart from that everything loaded fine).

However now it won't compile against the old Faust (2.50.6 or whatever it was) because the old API doesn't have the getWarningMessage() function. Is there a preprocessor macro in the faust headers that I can query for the version and conditionally include the getWarningMessage() function definition? I would love to not force everybody to update their Faust installation to keep using it.

sletz

Would FAUSTVERSION work for that?

See: https://github.com/grame-cncm/faust/blob/master-dev/architecture/faust/export.h#L28

Function added in 2.54.0, see: https://github.com/grame-cncm/faust/commit/a962231a9ca985cfa2ba513c01028872af62b115

BTW: now I understand why the latest published 2.54.9 version was also crashing !?

Christoph Hart

@sletz The problem is that this is a preprocessor string and it's impossible to use it in a conditional expression. I've added a manual projucer flag in the hi_faust_types module that the user has to switch on if he wants to keep using the old faust version (OLD_FAUST_API_SUPPORT), but this is only a very low level solution that only removes the getWarningMessage() function from the Faust wrappers and will probably cause issues again with the next breaking API change.

If you could add integer versions of the version, this would allow people to check the library version conditionally (and I could remove that flag and let the system figure it out automatically). Usually it's done like this:

#define FAUSTMAJORVERSION 2
#define FAUSTMINORVERSION 50
#define FAUSTPATCHVERSION 6

then I could use

#if  FAUSTMAJORVERSION <= 2 && FAUSTMINORVERSION <= 54
#define OLD_FAUST_API_SUPPORT 1
#else
#define OLD_FAUST_API_SUPPORT 0
#endif

sletz

OK, fixed in https://github.com/grame-cncm/faust/commit/fd8b6fa2496f55f3de0c531455f4cf9a8339586a.

sletz

Add classInit method in dsp_factory class, https://github.com/grame-cncm/faust/commit/90d40d425dac7557667387e3a8a5f4b44ce1ae43.

Version raised to 2.59.0, possible breaking for HISE integration.

sletz

Working with 2.59.1 without changes.

Lindon

@sletz said in Faust 2.58.11 to test:

Working with 2.59.1 without changes.

so are we saying 2.59.1 works with HISE?

sletz

Yes indeed.

Lindon

@Lindon said in Faust 2.58.11 to test:

@sletz said in Faust 2.58.11 to test:

Working with 2.59.1 without changes.

so are we saying 2.59.1 works with HISE?

Could someone please point me at Faust release 2.59.1 - the latest I can find is 2.54.9....

Lindon

for those looking....

https://github.com/DBraun/faust/actions/runs/4788344768

down the bottom of the page 2.59.2

Faust 2.58.11 to test

25

1.8k

10.4k

90.6k