Release date: 2026-09-24 / Chromium: 154.0.8037.57 / Language: 日本語 | English
The rendering engine is updated to Chromium 154.
Settings such as blink_web_view_set_javascript_enabled()
are now applied to the page.
Up to 1.2.2 the value was only stored: page scripts still ran
after passing FALSE.
| Function | 1.2.3 (measured on the shipped build) |
|---|---|
blink_web_view_set_javascript_enabled() |
Works (FALSE stops page scripts) |
blink_web_view_set_images_enabled() |
Works (FALSE stops image loading) |
blink_web_view_set_local_storage_enabled() |
Works (FALSE makes localStorage
unavailable) |
blink_web_view_set_default_font_size() |
Works (passing 30 gives 30px body text) |
blink_web_view_set_default_encoding() |
Not yet (pages without a declared charset stay
windows-1252) |
If your application must keep scripts from running on some pages,
update to this release.
set_default_encoding() will be investigated and handled in
a later release.
out_mime with parameters, such as
"text/html; charset=utf-8",<pre> text.fetch() on schemes listed in
BLINKGTK_EXTRA_SCHEMES. Up to 1.2.2 onlyapp / blinkgtk /
res / resource / ebook) allowed
itWithout blink_gtk_set_resources_path(), BlinkGTK now
also looks for icudtl.dat and the
other resources next to libblinkgtk. Up to 1.2.2 it only
looked next to the executable, so
applications built against the rpm / deb packages stopped with an ICU
error.
The search order is:
<libdir>/blinkgtk-0.1/chromium/ (rpm / deb)<libdir>/chromium/ (tarball)<libdir>/Applications that set the path keep working as before.
BlinkGTK no longer picks the Japanese input path itself; it
leaves the choice to GTK.
On Wayland this is GTK4's default text-input path, which is also what
the input method
side (IBus) recommends. On real Fedora 45 / GNOME hardware we confirmed,
with this path
selected, that conversion mode works and that closing views repeatedly
produces no input
method warnings (measured by selecting the same path through the switch
added in 1.2.2).
To get the 1.2.2 behaviour (naming IBus), set
BLINKGTK_IM_MODULE=ibus. It affects
BlinkGTK's views only. blink_gtk_get_im_module() returns
the value currently applied
(an empty string by default, meaning GTK decides).
blink_gtk_trace_mark("name") puts a marker in the trace.
The event is named after the
marker, the same shape as the web's performance.mark().
// Start with: --trace-startup=toplevel,blink.user_timing
blink_gtk_trace_mark("workload_ready");
/* ... the part to measure ... */
blink_gtk_trace_mark("idle_static_ready");Without the blink.user_timing category the call does
nothing.
text-decoration-inset — pull
underlines, overlines and line-throughs in from<iframe> —
the embedding page can size an iframe to its contentfont-width descriptor in
@font-face — an alias of
font-stretchThe feature list is Chromium 154's list on chromestatus.com. What we
measured ourselves
covers rendering, input and startup, not these typography features.
The workaround added in 1.2.1 (a vertical line breaking half-way
when a closing bracket
falls at the line end) is removed. Chromium 153 fixed
the same symptom upstream; we
kept the workaround in 1.2.2 after measuring that it no longer changed
the result
The header comment of the load-failed signal did not
match the implementation. The
correct handler signature is (the signal itself is unchanged):
static void on_load_failed (BlinkWebView *web_view, BlinkLoadError error_code,
const char *failing_uri, gpointer user_data);The header said blink_gtk_init_with_sandbox() was
not in the shared library yet, but it
has been since 1.2.2-build4. On Linux, passing NULL behaves
exactly like blink_gtk_init()
BLINKGTK_IM_MODULE=ibus restores the old behaviourFALSE will see page scripts or images stopset_javascript_enabled(FALSE) stops page scriptsfetch() and starting withoutblink_gtk_trace_mark() andperformance.mark() appear as events named after the
marker