The frame server could only capture by running a program once per frame
(grim). That works, but ~95% of a grim frame is waiting -- fork, connect,
handshake, round trip -- not encoding, and grim makes the compositor import
a *new* shared-memory buffer for every single frame.
This adds `--capture wayland[:OUTPUT]`: one long-lived zwlr-screencopy-v1
client per server, with one shm buffer reused for the whole run, capturing
inside the process. Measured here at 1280x720 (spike capture-speed, 200
frames; live re-check on Virtual-1, 60 frames):
grim (subprocess) 117.3 ms/frame 8.1 fps compositor 116.4 ms
grim-rs (Rust crate) 124.6 ms/frame 7.5 fps compositor 123.1 ms
this backend 78.0 ms/frame 13.9 fps compositor 67.5 ms
live, PNG both sides: this 128.5 ms/frame against grim's 389.3 ms (3.0x)
Three findings, each of which shaped the code:
* The win is buffer reuse, not language. grim and grim-rs both hand the
compositor a fresh buffer per frame, which is why their compositor CPU is
nearly double ours. The per-frame allocate-copy-swizzle that grim-rs pays
was measured in isolation at 2.4 ms of *client* time; the rest of its
deficit lands inside the compositor.
* ext-image-copy-capture-v1 is no faster (78.0 vs 77.7 ms) and its
create_frame path appears in a Hyprland segfault stack, so this keeps the
deprecated protocol grim uses. The switch is one constant when that crash
is fixed.
* PNG only, deliberately: PNG is the one encoding this package produces
without a dependency, so --format jpeg is refused rather than publishing a
Hello whose MIME type the frames do not have. A PNG frame is ~2.1 MB
against ~120 KB for JPEG q80, which is why the command backend stays the
default and why a libjpeg-turbo encoder (1.7 ms/frame, measured) is the
obvious follow-up.
The timeout is the load-bearing part of the client: pywayland only reads
events from read()/dispatch(block=True), neither of which takes a timeout,
and the server captures in a worker thread where SIGALRM is unavailable. It
therefore polls the display fd itself, so a compositor that stops answering
produces a CaptureError instead of a capture thread that never returns.
The protocol bindings are vendored under src/hhstream/_protocol/ exactly the
way hhstream_pointer already vendors its two, with regenerate.py recording
where each XML came from. No new dependency: pywayland was already required
by the input driver.
Contract: the new server flag and the two new module options are documented
in docs/CONTRACT.md, and docs/ARCHITECTURE.md's prediction that a native
screencopy client would be "a swap rather than a rewrite" is now the record
of what the swap cost.
Unit suite: 977 passed, 2 skipped, 0 failed (nix build -L .#checks.unit).
The live numbers were taken on this VM against the real compositor.
122 lines
4.6 KiB
Python
122 lines
4.6 KiB
Python
"""`hhstream.png`: the hand-rolled PNG encoder, checked against zlib itself.
|
|
|
|
There is no image library in the unit environment (the whole point of this
|
|
encoder is that it needs none), so the decoder below is written from the PNG
|
|
spec too. That is not a weakness of the test: it re-derives the scanlines from
|
|
the file the encoder produced and compares them with the pixels that went in,
|
|
which is exactly the property that matters -- a wrong stride, a wrong colour
|
|
type or a byte order slip all show up as a mismatch rather than as a file that
|
|
merely starts with the right eight bytes.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import struct
|
|
import zlib
|
|
|
|
import pytest
|
|
|
|
from hhstream.png import encode_rgb
|
|
|
|
SIGNATURE = b"\x89PNG\r\n\x1a\n"
|
|
|
|
|
|
def decode(png: bytes) -> tuple[int, int, bytes]:
|
|
"""Return (width, height, rgb scanlines) for a PNG this encoder can produce."""
|
|
assert png.startswith(SIGNATURE), "missing PNG signature"
|
|
offset = len(SIGNATURE)
|
|
width = height = None
|
|
idat = b""
|
|
while offset < len(png):
|
|
(length,) = struct.unpack(">I", png[offset:offset + 4])
|
|
kind = png[offset + 4:offset + 8]
|
|
payload = png[offset + 8:offset + 8 + length]
|
|
crc = png[offset + 8 + length:offset + 12 + length]
|
|
assert struct.unpack(">I", crc)[0] == zlib.crc32(kind + payload) & 0xFFFFFFFF, (
|
|
f"bad CRC on {kind!r} chunk -- the file is corrupt"
|
|
)
|
|
if kind == b"IHDR":
|
|
width, height, depth, color, comp, filt, interlace = struct.unpack(">IIBBBBB", payload)
|
|
assert (depth, color, comp, filt, interlace) == (8, 2, 0, 0, 0)
|
|
elif kind == b"IDAT":
|
|
idat += payload
|
|
offset += 12 + length
|
|
|
|
assert width is not None and height is not None, "no IHDR chunk in the file"
|
|
raw = zlib.decompress(idat)
|
|
stride = width * 3
|
|
assert len(raw) == height * (1 + stride), "the scanline payload is the wrong length"
|
|
out = bytearray()
|
|
for row in range(height):
|
|
start = row * (1 + stride)
|
|
assert raw[start] == 0, "every scanline should be written unfiltered"
|
|
out += raw[start + 1:start + 1 + stride]
|
|
return width, height, bytes(out)
|
|
|
|
|
|
def bgrx(width: int, height: int, stride: int | None = None) -> bytes:
|
|
"""A frame in the compositor's memory order: B, G, R, X per pixel."""
|
|
stride = stride if stride is not None else width * 4
|
|
out = bytearray()
|
|
for y in range(height):
|
|
row = bytearray()
|
|
for x in range(width):
|
|
row += bytes(((x * 3) % 256, (y * 5) % 256, (x + y) % 256, 0x00))
|
|
row += b"\xee" * (stride - width * 4) # row padding, if any
|
|
out += row
|
|
return bytes(out)
|
|
|
|
|
|
def test_pixels_round_trip_in_the_right_channel_order():
|
|
frame = bgrx(4, 2)
|
|
width, height, rgb = decode(encode_rgb(4, 2, 16, frame))
|
|
assert (width, height) == (4, 2)
|
|
# Pixel (x, y) went in as B,G,R,X; it must come out as R,G,B.
|
|
for y in range(2):
|
|
for x in range(4):
|
|
b, g, r, _ = frame[(y * 4 + x) * 4:(y * 4 + x) * 4 + 4]
|
|
assert rgb[(y * 4 + x) * 3:(y * 4 + x) * 3 + 3] == bytes((r, g, b)), (
|
|
f"channel order is wrong at ({x}, {y})"
|
|
)
|
|
|
|
|
|
def test_a_padded_stride_is_honoured_rather_than_assumed():
|
|
"""`stride` is the compositor's row pitch, not `width * 4`.
|
|
|
|
Assuming it would shear the image, which is the bug grim 1.5 shipped -- so
|
|
the padding bytes below are deliberately not black: if they leak into the
|
|
output the comparison fails.
|
|
"""
|
|
width, height, stride = 3, 2, 20 # 12 meaningful bytes, 8 bytes of padding
|
|
frame = bgrx(width, height, stride)
|
|
_, _, rgb = decode(encode_rgb(width, height, stride, frame))
|
|
assert rgb == b"".join(
|
|
bytes(reversed(frame[(y * stride + x * 4):(y * stride + x * 4) + 3]))
|
|
for y in range(height)
|
|
for x in range(width)
|
|
)
|
|
|
|
|
|
def test_a_memoryview_of_a_mapping_is_accepted():
|
|
"""The caller passes `memoryview(shm_map)`, not `bytes`."""
|
|
frame = bgrx(2, 2)
|
|
assert encode_rgb(2, 2, 8, memoryview(frame)) == encode_rgb(2, 2, 8, frame)
|
|
|
|
|
|
@pytest.mark.parametrize(
|
|
"kwargs, message",
|
|
[
|
|
(dict(width=0, height=4, stride=0), "positive size"),
|
|
(dict(width=4, height=0, stride=16), "positive size"),
|
|
(dict(width=4, height=4, stride=8), "too small"),
|
|
],
|
|
)
|
|
def test_bad_geometry_is_rejected_before_anything_is_encoded(kwargs, message):
|
|
with pytest.raises(ValueError, match=message):
|
|
encode_rgb(pixels=b"\x00" * 64, **kwargs)
|
|
|
|
|
|
def test_a_short_buffer_is_rejected_rather_than_encoding_garbage():
|
|
with pytest.raises(ValueError, match="needs"):
|
|
encode_rgb(4, 4, 16, b"\x00" * 63)
|