[02:53] <RAOF> Oooooh, yeah!
[02:53] <RAOF> the_buffer_allocator() exists!
[07:30] <RAOF> Thanks, past-me. renderer-egl-generic is very handy.
[07:35] <RAOF> OK. So, the probing doesn't work because we need to ensure all the GLRenderingProviders can import client buffers from whatever has allocated them, and that's not necessarily the same platform. Fair enough.
[07:35] <RAOF> 🤔 But why didn't the dynamic_cast fail and cause Mir to crash?
[07:40] <RAOF> Ah, right. Because the underlying buffer type isn't platform specific.
[08:05] <RAOF> Hm, but no. The simplest thing that fixes this is to move the probe earlier, so that the GBM renderer doesn't load where it will be useless.
[08:10] <RAOF> Today:... (full message at <https://libera.ems.host/_matrix/media/v3/download/libera.chat/86efe3aa8456550b2e0c8db3fe4d0faec2e9ebf8>)
[08:11] <RAOF>  * Today:... (full message at <https://libera.ems.host/_matrix/media/v3/download/libera.chat/77b0f432911af4ae6c09403bfcc688d65c11703c>)
[08:57] <alan_g[m]> I hate it when the graphics-core22 documentation says one thing but the mesa-core22 implementation does another and users rely on the latter.
[08:57] <alan_g[m]> ...and when mesa-core22/beta matches the documentation (why doesn't that break mir-test-tools?)
[08:58] <alan_g[m]> ...and when the most recent person to touch all this is away for the week
[09:45] <alan_g[m]> ...looking closer, the mesa-core22/beta didn't change what actually matters and the documentation is clearly out of step. (Fixed)
[17:12] <alan_g[m]> Today:
[17:12] <alan_g[m]> * Investigating mismatch between graphics-core22 docs and implementation
[17:12] <alan_g[m]> * Planning next piece of DnD implementation (finally, I have some clarity on implementation approach)