fix: dpms cycle causes blurred wallpaper

[Anaethelion]

When DPMS turns off and back on, the EGL context is recreated via check_context but no configure event follows. EglContext::new always created a 10x10 placeholder surface, so the wallpaper rendered at that size and was stretched — appearing blurred.

Fix: Use real display dimensions in EglContext::new when known. Assign the context before drawing and remove frame callback requests from error paths to prevent a tight error loop.

For context, here is the backtrace triggered by cycling dpms off/on:

ERROR [wpaperd::surface]
   0: Failed to draw on display DP-1
   1: Failed to draw the wallpaper
   2: Failed to swap EGL buffers
   3: Failed to draw the content of the GL buffer
   4: An Surface argument does not name a valid surface (window, pixel buffer or pixmap) configured for GL rendering.

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ BACKTRACE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
                                ⋮ 3 frames hidden ⋮
   4: wpaperd::render::egl_context::EglContext::draw::h2cfd419ad98f7a10
      at <unknown source file>:<unknown line>
   5: wpaperd::surface::Surface::try_drawing::h13784a2ae9118c8c
      at <unknown source file>:<unknown line>
   6: <core::slice::iter::IterMut<T> as core::iter::traits::iterator::Iterator>::for_each::hac05d0588b70c788
      at <unknown source file>:<unknown line>
   7: wpaperd::main::h2d7449a618b4ff0d
      at <unknown source file>:<unknown line>
   8: std::sys::backtrace::__rust_begin_short_backtrace::hddf1794ab0d6f4f6
      at <unknown source file>:<unknown line>
   9: std::rt::lang_start::{{closure}}::h1492e302212c8192
      at <unknown source file>:<unknown line>
  10: std::rt::lang_start_internal::hb84cc625940d332a
      at <unknown source file>:<unknown line>
  11: main<unknown>
      at <unknown source file>:<unknown line>
  12: __libc_start_main<unknown>
      at <unknown source file>:<unknown line>
  13: _start<unknown>
      at <unknown source file>:<unknown line>

Tested on Hyprland 0.54.1 built from branch v0.54.1 at commit 4b07770b9ef1cceb2e6f56d33538aaffb9186b9c

---

This fix was developed with AI assistance (Claude), feel free to disregard and close.

[danyspin97]

Hello @Anaethelion and thank you for the PR! I now get the bug on NVIDIA drivers, but I am not sure about some changes, I'll comment more on the actual code. On a sidenote, there is also #177 that attempt to fix the same thing. Can you please give it a try and let me know if it works the same?

[danyspin97]

Why are we resizing after creating a surface with the correct size?

[danyspin97]

This only schedule the compositor to call frame callback on this surface the next time it needs to be drawn. So calling it here shouldn't lead to a loop, I guess.

[danyspin97]

This is a cosmetic change, not needed, imho.

[danyspin97]

I think we need both this PR and #177, as they fix different things.