do not apply storage attributes to member of struct

[battlesnake]

Describe the PR
Compilation fails when audio interrupt-endpoint is enabled, due to storage attributes on a struct member.
I removed the attributes.

Additional context

[hathach]

it does not fail with my compiling. This is required for some mcu. Could you provide info on your setup

[battlesnake]

This is using GCC 12.2.0, st/synopsys port, STM32H7.

The attributes (CFG_TUSB_MEM_SECTION CFG_TUSB_MEM_ALIGN) are being applied to a member of a struct:

struct {
  ...
  CFG_TUSB_MEM_SECTION CFG_TUSB_MEM_ALIGN uint8_t ep_int_ctr_buf[CFG_TUD_AUDIO_INT_CTR_EP_IN_SW_BUFFER_SIZE];
  ...
}

I don't think it makes sense (regarding memory layout) for some members of a struct to be in a different linker section to the rest of the struct? GCC certainly doesn't like this.

The only instance of the struct is (around line 413):

CFG_TUSB_MEM_SECTION audiod_function_t _audiod_fct[CFG_TUD_AUDIO];

So the struct (and therefore its members) are already in the section denoted by CFG_TUSB_MEM_SECTION.
The alignment is a separate issue, perhaps that attribute should stay.

Setting CFG_TUD_AUDIO_INT_CTR_EPSIZE_IN to non-zero (e.g. 6) enables that struct member.
If you have CFG_TUSB_MEM_SECTION defined with a specific section, then this combination should be sufficient to trigger a compiler error.

For configuration, I have:

#define _d1ram_bss __attribute__((__section__(".ram_d1.bss")))
...
#define dma_align_bytes (32)
#define _dma_align __attribute__((__aligned__(dma_align_bytes)))
...
#define CFG_TUSB_MEM_SECTION _d1ram_bss
#define CFG_TUSB_MEM_ALIGN _dma_align

#define CFG_TUD_AUDIO_INT_CTR_EPSIZE_IN 6

Which, expanded, is:

#define CFG_TUSB_MEM_SECTION __attribute__((__section__(".ram_d1.bss")))
#define CFG_TUSB_MEM_ALIGN __attribute__((__aligned__(32)))

#define CFG_TUD_AUDIO_INT_CTR_EPSIZE_IN 6

[battlesnake]

Currently, the ep_int_ctr member is used, but never assigned.

So I'm guessing that I'm the first person to try building a UAC2 configuration with interrupt/status endpoint enabled, and the feature currently isn't fully implemented in tinyusb yet.

I'm trying to implement it, but not sure if I'll succeed.

[hathach]

I don't think it makes sense (regarding memory layout) for some members of a struct to be in a different linker section to the rest of the struct? GCC certainly doesn't like this.

Oh, this is right, please remove the CFG_TUSB_MEM_SECTION but keep the CFG_TUSB_MEM_ALIGN

[battlesnake]

Done

[hathach]

perfect, thank you