Arch Linux User Repository

Another issue with GCC 14 being stricter than whatever the devs regularly use. Patched.

Looks like vgmstream123.c forgot to include wav_utils.h from its own directory:

vgmstream123.c: In function ‘play_vgmstream’:
vgmstream123.c:399:13: error: implicit declaration of function ‘swap_samples_le’ [-Wimplicit-function-declaration]
  399 |             swap_samples_le(buffer, output_channels * to_do, 0);
      |             ^~~~~~~~~~~~~~~
make[1]: *** [Makefile.autotools:447: vgmstream123.o] Error 1

A new required header file was added. Updated install-headers.patch to include it.

makepkg check step now fails:

In file included from <stdin>:1:
/build/vgmstream-git/src/test/usr/include/vgmstream/vgmstream.h:22:10: fatal error: util/vgmstream_limits.h: No such file or directory
   22 | #include "util/vgmstream_limits.h"
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
==> ERROR: A failure occurred in check().
    Aborting...

Indeed, that code is straight-up broken on 64-bit platforms and GCC is right to reject it. Pushed a patch to circumvent the issue, but the fix should really be upstreamed, which I may get around to later this week.

Now upstream is failing to build due to new warnings in GCC 14, due to passing a uint32_t by pointer to a parameter asking for a pointer to size_t.

In file included from meta/ubi_hx.c:5:
meta/../util/chunks.h:38:127: note: expected 'size_t *' {aka 'long unsigned int *'} but argument is of type 'uint32_t *' {aka 'unsigned int *'}
   38 | int find_chunk_riff_ve(STREAMFILE* sf, uint32_t chunk_id, off_t start_offset, size_t max_size, off_t* p_chunk_offset, size_t* p_chunk_size, int big_endian);
      |                                                                                                                       ~~~~~~~~^~~~~~~~~~~~
meta/ubi_hx.c:396:103: error: passing argument 6 of 'find_chunk_riff_ve' from incompatible pointer type [-Wincompatible-pointer-types]
  396 |             if (!find_chunk_riff_ve(sf, 0x61746164,riff_offset + 0x0c,riff_size - 0x0c, &chunk_offset,&chunk_size, hx->big_endian)) {
      |                                                                                                       ^~~~~~~~~~~
      |                                                                                                       |
      |                                                                                                       uint32_t * {aka unsigned int *}
meta/../util/chunks.h:38:127: note: expected 'size_t *' {aka 'long unsigned int *'} but argument is of type 'uint32_t *' {aka 'unsigned int *'}
   38 | int find_chunk_riff_ve(STREAMFILE* sf, uint32_t chunk_id, off_t start_offset, size_t max_size, off_t* p_chunk_offset, size_t* p_chunk_size, int big_endian);
      |                                                                                                                       ~~~~~~~~^~~~~~~~~~~~
make[2]: *** [Makefile.autotools:2704: meta/ubi_hx.lo] Error 1

Alright, well it seems that your plugin does need the api.h so I added it here.

@Deewiant: I'm not really involved with the project any more. Though I do occasionally use it still. I don't know that the api.h was really needed by a plugin, I may be able to get away without it.

I made a deadbeef plugin, it requires this AUR package to build, and links to it dynamically.

https://bitbucket.org/losnoco/deadbeef_plugins - under plugins/vgm

@kode54 You might know something I don't but as far as I can tell those aren't really used, at least yet. I've based the headers to install on delivering a working vgmstream.h, which seems like the "entry point" to the public API. api.h currently doesn't do anything other than include base/plugins.h, which was apparently moved from plugins.h recently but that one has never been installed here either.

I guess the problem is that this thing isn't really a library in the sense of being intended for use by other projects. Or if it is, I don't understand where the line is between public API and internals.

(As a side note I tried to do a CMake conversion of the PKGBUILD but that one apparently doesn't install a shared library at all and just statically links both of the binaries, so it's in even worse shape from my perspective...)

New headers for the install headers patch:

api.h
base/plugins.h