libglvnd: fix build with newer clang

[reckenrode]

Description of changes

This was discovered while working on #229786. This change silences new warnings regarding integer to pointer conversions. Targeting staging due to the number of rebuilds.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 23.05 Release Notes (or backporting 22.11 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

[primeos]

Thanks. One note and a question:

This was discovered while working on #229786. This change silences new warnings regarding integer to pointer conversions. Targeting staging due to the number of rebuilds.

Would've been lovely to have that information in the Git commit message as well (so that it's properly documented in the Git history).

Is the Git commit message subject correct? Does this really fix the build or does it only silence some noisy warnings (as I'd expect from the diff)?

[reckenrode]

Would've been lovely to have that information in the Git commit message as well (so that it's properly documented in the Git history).

I’ll keep that in mind for the next round of PRs fixing other breakage, but shouldn’t the commit be linked back to the PR by GitHub?

Is the Git commit message subject correct? Does this really fix the build or does it only silence some noisy warnings (as I'd expect from the diff)?

It fails with the following errors, which the additional flag suppresses.

libegl.c:257:38: error: incompatible integer to pointer conversion passing 'EGLNativeDisplayType' (aka 'int') to parameter of type 'void *' [-Wint-conversion]
    if (gbmSupported && IsGbmDisplay(display_id)) {
                                     ^~~~~~~~~~
libegl.c:161:38: note: passing argument to parameter 'native_display' here
static EGLBoolean IsGbmDisplay(void *native_display)
                                     ^
libegl.c:260:46: error: incompatible integer to pointer conversion passing 'EGLNativeDisplayType' (aka 'int') to parameter of type 'void *' [-Wint-conversion]
    if (waylandSupported && IsWaylandDisplay(display_id)) {
                                             ^~~~~~~~~~
libegl.c:201:42: note: passing argument to parameter 'native_display' here
static EGLBoolean IsWaylandDisplay(void *native_display)
                                         ^
libegl.c:263:38: error: incompatible integer to pointer conversion passing 'EGLNativeDisplayType' (aka 'int') to parameter of type 'void *' [-Wint-conversion]
    if (x11Supported && IsX11Display(display_id)) {
                                     ^~~~~~~~~~
libegl.c:177:38: note: passing argument to parameter 'dpy' here
static EGLBoolean IsX11Display(void *dpy)
                                     ^
libegl.c:389:55: error: incompatible integer to pointer conversion passing 'EGLNativeDisplayType' (aka 'int') to parameter of type 'void *' [-Wint-conversion]
            return GetPlatformDisplayCommon(platform, display_id, NULL, "eglGetDisplay");
                                                      ^~~~~~~~~~
libegl.c:271:15: note: passing argument to parameter 'native_display' here
        void *native_display, const EGLAttrib *attrib_list,
              ^
libegl.c:396:51: error: incompatible integer to pointer conversion passing 'EGLNativeDisplayType' (aka 'int') to parameter of type 'void *' [-Wint-conversion]
        return GetPlatformDisplayCommon(EGL_NONE, display_id, NULL, "eglGetDisplay");
                                                  ^~~~~~~~~~
libegl.c:271:15: note: passing argument to parameter 'native_display' here
        void *native_display, const EGLAttrib *attrib_list,
              ^
libegl.c:405:47: error: incompatible integer to pointer conversion passing 'EGLNativeDisplayType' (aka 'int') to parameter of type 'void *' [-Wint-conversion]
    return GetPlatformDisplayCommon(platform, display_id, NULL, "eglGetDisplay");
                                              ^~~~~~~~~~
libegl.c:271:15: note: passing argument to parameter 'native_display' here
        void *native_display, const EGLAttrib *attrib_list,
              ^

[primeos]

I’ll keep that in mind for the next round of PRs fixing other breakage, but shouldn’t the commit be linked back to the PR by GitHub?

Yes, but that link isn't part of the Git repository (information that could get lost, etc.) and then one has to invest extra time for looking everything up via GitHub - which is very inconvenient for Git power users.

It fails with the following errors, which the additional flag suppresses.

Thanks :)
Strange that the -Wno-error flag doesn't seem to apply here...