This is an automated email from the git hooks/post-receive script.
git pushed a commit to branch wl/browser-all
in repository enlightenment.
View the commit online.
commit 2ff8856c958d5c73fb68a27dd3ad847b3c6bc184
Author: Cedric BAIL <[email protected]>
AuthorDate: Mon Aug 17 19:06:44 2026 -0600
tests - let the browser tier run on the GPU, and say so when it does not
--disable-gpu has been on every chromium and brave run since the tier was
written, and the default stays that way: the suite's canvas is software unless
a test asks for E_TEST_ENGINE=gl, and a browser told to use a GPU against a
compositor that cannot import its buffers spends its startup failing over to
software, slowly and differently each time.
What was wrong was the reading of it. Task #24 recorded "Brave runs on GPU
buffers" and that was never true here. Measured with WAYLAND_DEBUG on the
default configuration: brave binds zwp_linux_dmabuf_v1 at version 4, gets the
whole feedback exchange - 2048-byte format table, main_device, one tranche -
destroys the feedback object, and then allocates every buffer it attaches from
wl_shm. Six shm buffers, zero dmabuf.
With E_TEST_BROWSER_GPU=1 on E_TEST_BACKEND=x11 E_TEST_ENGINE=gl it is six
dmabuf buffers through create_params/create_immed and one shm, no protocol
error on either side, and the full maximise/move/resize/second-screen sequence
passes. That is the first end-to-end evidence that E's dmabuf import works
with a client nobody here wrote.
Co-Authored-By: Claude Opus 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FtoiXoSKUmZb6Aix6U3GZS
---
src/tests/wayland/browser-run.sh | 18 +++++++++++++++++-
1 file changed, 17 insertions(+), 1 deletion(-)
diff --git a/src/tests/wayland/browser-run.sh b/src/tests/wayland/browser-run.sh
index f26c95d52..1987ba1a3 100755
--- a/src/tests/wayland/browser-run.sh
+++ b/src/tests/wayland/browser-run.sh
@@ -128,6 +128,22 @@ JS
COMMON="$COMMON MOZ_ENABLE_WAYLAND=1 MOZ_CRASHREPORTER_DISABLE=1"
;;
chromium|brave)
+ # --disable-gpu unless E_TEST_BROWSER_GPU says otherwise, and the default
+ # is the conservative one rather than the right one. The suite's canvas is
+ # software unless a test asks for E_TEST_ENGINE=gl, and a browser told to
+ # use the GPU against a compositor that cannot import its buffers spends
+ # its startup failing over to software - slowly, and differently each time.
+ #
+ # It does mean the default configuration says nothing about the GPU path.
+ # A browser with this flag still binds zwp_linux_dmabuf_v1 and still reads
+ # the whole feedback exchange; it simply never allocates from it, and every
+ # buffer it attaches comes from wl_shm. Anyone reading a green browser run
+ # as evidence about dmabuf is reading it wrong - that was recorded once
+ # already, as task #24.
+ GPU_FLAGS="--disable-gpu"
+ if [ "${E_TEST_BROWSER_GPU:-0}" = "1" ]; then
+ GPU_FLAGS=""
+ fi
# Quote --host-resolver-rules. Unquoted, the shell globs the * against the
# working directory and the browser starts with no rule at all.
set -- "$BIN" \
@@ -137,7 +153,7 @@ chromium|brave)
--password-store=basic --use-mock-keychain \
--disable-component-update --disable-background-networking \
--host-resolver-rules="MAP * ~NOTFOUND" \
- --ozone-platform=wayland --disable-gpu \
+ --ozone-platform=wayland $GPU_FLAGS \
"$URL"
;;
*)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
