- Apr 22, 2013
-
-
Colin Walters authored
-
Colin Walters authored
Fixes the test suite. https://bugzilla.gnome.org/show_bug.cgi?id=697436
-
- Apr 05, 2013
-
-
Ray Strode authored
There are a number of cases where toggle notifies run that we can get away with handling them directly. This commit does that, to minimize the number of idle functions getting queued. https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit 0adcb39d)
-
Ray Strode authored
We don't want to be mucking with javascript from two threads at once. This commit defers toggle notification processing to the main thread, and introduces a mutex to prevent the garbage collector and toggle notifications from running at the same time. https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit 50a3d867)
-
Ray Strode authored
Right now we only set the GC callback if gc-notifications are enabled. This commit makes us always set the callback as prep work toward fixing a GC deadlock. https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit 7ab5e0e4)
-
Ray Strode authored
There's a small race where the GC notification idle id might get zeroed before the idle runs, if the GC callback is run from a non-main thread. That means there's the potential for the GC notification idle to get dispatched after the context is finalized. This commit fixes that by protecting access to the idle id with a lock. https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit d112a716)
-
Ray Strode authored
This commit fixes a typo that can only see when compiling gjs with extra debugging. https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit e2d80eb9)
-
Ray Strode authored
It's possible for a closure to have 0 arguments. When this happens we end up calling alloca(0) which is undefined. This commit protects against argc == 0. (concretely, I was seeing this happen from g_closure_invoke(clsoure, &result_value, 0, NULL, NULL) in glib's gsourceclosure.c: source_closure_callback function) https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit a026cc5e)
-
Ray Strode authored
There are a few early return cases where a JS_BeginRequest isn't getting paired with JS_EndRequest. This commit fixes those cases. https://bugzilla.gnome.org/show_bug.cgi?id=670200 (cherry picked from commit 27a13959)
-
Giovanni Campagna authored
It would cast a jsval of the wrong type to object and pass NULL to gjs_gtype_get_actual_gtype() Tests included. https://bugzilla.gnome.org/show_bug.cgi?id=696933 (cherry picked from commit f4a5b2c9)
-
- Mar 26, 2013
-
-
Colin Walters authored
-
-
Colin Walters authored
-
Upstream are moving toward each runtime only having a single context. As such all GC functions (except JS_MaybeGC), now take runtime as the argument, rather than the context. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
Both these api are deprecated and removed. For now add wrappers to compat.h that provide equivalent functionality. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
These calls to localeCompare are supposed to fail, however they cause the testcase to fail with 'function() undefined'. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
For some reason the setter fails to get called when JSPROP_SHARED is set. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
E4X is disabled by default in esr17, re-enable this. jscntxt.h is no longer a public header and replace some deprecated API, with equivalent calls. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
JS_FinalizeStub no longer exists. JSCLASS_MARK_IS_TRACE flag is no longer required. JSClass prototype has some changes to the optional members, xdrObject and hasInstance have been removed, reserved has been moved to the end of the prototype. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
All functions which are callbacks for JSClass, now have new types which wraps the Objects in an extra pointer. We need to cast and then derefence these objects. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
JS_SetPrivate/JS_GetPrivate/JS_GetPrototype no longer require context as first arg. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
JS_GetClass no longer takes *context as first argument, consequently the JS_GET_CLASS Macro is also removed. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
This now requires a JSBool argument specifying UTF-8 support. https://bugzilla.gnome.org/show_bug.cgi?id=690982
-
- Mar 21, 2013
-
-
Colin Walters authored
-
- Mar 13, 2013
-
-
- Mar 11, 2013
-
-
Missing bit of 72afb882
-
- Mar 07, 2013
-
-
-
Jasper St. Pierre authored
Now that all modules are guaranteed to be static and toplevel, we can remove a large set of code that's been around simply for support for nested native modules. https://bugzilla.gnome.org/show_bug.cgi?id=694873
-
Jasper St. Pierre authored
Rather than picking it up indirectly through meta properties. https://bugzilla.gnome.org/show_bug.cgi?id=694873
-
-
-
Jasper St. Pierre authored
Now that we have significantly less of them, there's no real reason to not link in the native implementations of this... https://bugzilla.gnome.org/show_bug.cgi?id=694873
-
- Feb 28, 2013
-
-
-
-
Jasper St. Pierre authored
Rather than use g_once_init_enter, simply use the standard place for the one-time initialization sequence. https://bugzilla.gnome.org/show_bug.cgi?id=694828
-
- Feb 26, 2013
-
-
Colin Walters authored
Had this locally but failed to rebase it...
-
-
David King authored
Add the _GI_DISABLE_CAIRO define to the gobject-introspection scanner flags for the Regress-1.0 GIR. Require cairo-gobject if cairo support is enabled. https://bugzilla.gnome.org/show_bug.cgi?id=694373
-