test(pixelflux): Pin the wire header in a Rust test, where a Python suite drove the bindings
tests/integration/test_wire_header_picture_type.py captured through
pixelflux's bindings to check what each stripe's header says: the type
byte, the IDR nibble with the codec beside it, and the payload starting
where the header ends (an Annex-B start code for H.264, SOI for JPEG),
on X11 and on the in-process compositor, with the first picture an IDR,
the next ones not, and a requested keyframe arriving as one.
crates/pixelflux/tests/wire_header.rs makes the same checks through the
Rust API, on an Xvfb of its own (repainted through x11rb where JPEG's
damage needs it) and on the compositor, one per process, so a binary of
its own, ignored where cargo runs every test at once. CI's native
capture job installs xvfb and runs it, on x86_64 and aarch64 where the
suite ran on x86_64 alone. The suite goes.
Here: 31/31 checks, the suite's set (x11 hardware path, x11 CPU H.264,
x11 JPEG, Wayland both H.264 paths), 17.6 s on four cores.