Wayland: Commit surface on window creation

[Riteo]

Fixes #94204.
Possible regression from #93684.

This is needed for initializing xdg_surfaces when not using libdecor. Now the pure xdg_shell code path should work again.

[em-s-h]

Hi, I tried to compile this to test if #94204 had been fixed but I got this error during compilation.

[Initial build] Compiling thirdparty/glad/gl.c ...
In file included from thirdparty/vulkan/vk_mem_alloc.cpp:7:
thirdparty/vulkan/vk_mem_alloc.h: In function 'void VmaUint32ToStr(char*, size_t, uint32_t)':
thirdparty/vulkan/vk_mem_alloc.h:2821:9: error: 'snprintf' was not declared in this scope
 2821 |         snprintf(outStr, strLen, "%u", static_cast<unsigned int>(num));
      |         ^~~~~~~~
thirdparty/vulkan/vk_mem_alloc.h:2655:1: note: 'snprintf' is defined in header '<cstdio>'; this is probably fixable by adding '#include <cstdio>'
 2654 |     #include <mutex>
  +++ |+#include <cstdio>
 2655 | #else
thirdparty/vulkan/vk_mem_alloc.h: In function 'void VmaUint64ToStr(char*, size_t, uint64_t)':
thirdparty/vulkan/vk_mem_alloc.h:2825:9: error: 'snprintf' was not declared in this scope
 2825 |         snprintf(outStr, strLen, "%llu", static_cast<unsigned long long>(num));
      |         ^~~~~~~~
thirdparty/vulkan/vk_mem_alloc.h:2825:9: note: 'snprintf' is defined in header '<cstdio>'; this is probably fixable by adding '#include <cstdio>'
thirdparty/vulkan/vk_mem_alloc.h: In function 'void VmaPtrToStr(char*, size_t, const void*)':
thirdparty/vulkan/vk_mem_alloc.h:2829:9: error: 'snprintf' was not declared in this scope
 2829 |         snprintf(outStr, strLen, "%p", ptr);
      |         ^~~~~~~~
thirdparty/vulkan/vk_mem_alloc.h:2829:9: note: 'snprintf' is defined in header '<cstdio>'; this is probably fixable by adding '#include <cstdio>'
scons: *** [thirdparty/vulkan/vk_mem_alloc.linuxbsd.editor.x86_64.o] Error 1
In file included from ./core/os/thread.h:42,
                 from drivers/alsa/audio_driver_alsa.h:37,
                 from drivers/alsa/audio_driver_alsa.cpp:31:
./core/templates/safe_refcount.h:135:49: warning: template-id not allowed for constructor in C++20 [-Wtemplate-id-cdtor]
  135 |         _ALWAYS_INLINE_ explicit SafeNumeric<T>(T p_value = static_cast<T>(0)) {
      |                                                 ^
./core/templates/safe_refcount.h:135:49: note: remove the '< >'
In file included from ./core/os/thread.h:42,
                 from drivers/pulseaudio/audio_driver_pulseaudio.h:37,
                 from drivers/pulseaudio/audio_driver_pulseaudio.cpp:31:
./core/templates/safe_refcount.h:135:49: warning: template-id not allowed for constructor in C++20 [-Wtemplate-id-cdtor]
  135 |         _ALWAYS_INLINE_ explicit SafeNumeric<T>(T p_value = static_cast<T>(0)) {
      |                                                 ^
./core/templates/safe_refcount.h:135:49: note: remove the '< >'
scons: building terminated because of errors.
[Time elapsed: 00:20:59.712]

[Riteo]

@em-s-h that's quite baffling, the patch is a function call and an empty line...

Have you cherry-picked the change or something like that? Could you try doing a hard reset to this branch and doing a git clean -fdx (NOTE: it will delete everything "foreign" to the repo, including custom.py!)

[em-s-h]

Have you cherry-picked the change or something like that?

No, I simply cloned your fork and tried to build using the aur PKGBUILD script, although after reading the Godot documentation on compiling for linux it says that LTO (the default on the script) requires 7GB of ram while I only have 4, so could that cause this issue?

[Riteo]

@em-s-h The PKGBUILD will very probably try building the master branch of my fork, which is terribly outdated. You will have to switch to my branch (committed-to-the-cause), although I recommend to just build it manually, which isn't that hard.

Supposing you have all dependencies, once switched into the branch, a plain scons invocation should work fine.

Regarding the LTO thing, yeah 4 GBs is nothing, as it's extremely resource intensive and, since you're just testing a bugfix, I would not bother using it at all.

[BasilYes]

Patch work for me

[em-s-h]

I can also confirm that the patch works.

[Riteo]

Thank you folks for testing!

[akien-mga]

Thanks!