Repository navigation
onBeforeContextMenu never fires for a synthetic OSR right-click (root cause not yet isolated) #17
Description
Activity
- added a commit that references this issue
on Aug 29, 2026 Root-caused and fixed -- this was never a JCEF/CEF bug, just three compounding test-technique gaps in how the synthetic right-click was constructed:
- Wrong modifier field: the synthetic
MouseEvent'smodifiersconstructor argument was0.native/CefBrowser_N.cpp'sGetCefModifiers()readsgetModifiersEx()(the extended modifier mask), not the deprecatedgetModifiers()-- andMouseEvent's legacy constructor does NOT auto-populategetModifiersEx()from thebuttonargument alone. Confirmed via a standalone repro. Fix: passInputEvent.BUTTON3_DOWN_MASK. - Missing focus: real interactive use gives the OSR canvas AWT input focus before any click reaches it --
CefBrowserOsrregisters aFocusListenerthat callsbrowser.setFocus(true)onFOCUS_GAINED, which reaches CEF's own internalWebContentsfocus. Confirmed by reading CEF's own reference OSR implementation (tests/cefclient/browser/browser_window_osr_gtk.cc'sClickEvent(), which callsgtk_widget_grab_focus()on every mouse-down).Component.dispatchEvent()bypasses AWT's focus-transfer machinery entirely, sobrowser.setFocus(true)must be called explicitly. - Dispatched too early: the click was fired immediately in
onLoadingStateChange, before the OSR GL surface had a chance to realize/paint. A short delay (500ms) before dispatch was needed.
With all three fixed,
onBeforeContextMenufires reliably (verified across many isolated runs).CefMenuModel.clear()is called inside the callback to suppress the real native popup CEF would otherwise show, avoiding a genuine "modal UI blocks teardown" hang risk (same class as #12).This unblocked the single largest remaining 0%-covered code in the fork:
CefMenuModel_N.cppwent 3% → 84%,CefContextMenuParams_N.cppwent 13% → 79%,context_menu_handler.cppwent 0% → 62%. Total native coverage jumped 48.2% → 55% from this one fix. SeeCefContextMenuTest.javain the fork for the working repro/technique.- Wrong modifier field: the synthetic
- added 6 commits that reference this issue
on Aug 30, 2026
Summary
CefContextMenuHandler.onBeforeContextMenunever fires in response to a synthetic right-click dispatched to an OSR browser's UI component -- confirmed on this fork's current CEF version (146.0.10+g8219561+chromium-146.0.7680.179).Root cause not yet isolated -- this issue is filed to keep a reproducer on record (per this fork's practice of tracking real findings, even partial ones) rather than because the exact mechanism is understood.
Repro
Added
java/tests/junittests/CefContextMenuTest.java(currently@Disabledwith a link to this issue -- remove@Disabledto re-run it):CefBrowserOsr'sGLCanvasdoes register a realMouseListenerthat forwards every AWTMouseEventtosendMouseEvent()/native CEF (confirmed by readingCefBrowserOsr.java), so the mechanism this test relies on is real.BUTTON3press+releaseMouseEventviaComponent.dispatchEvent()tobrowser.getUIComponent(), right afteronLoadingStateChangereports the page finished loading.onBeforeContextMenufires with non-nullparams/model. Actual: it never fires -- the test hangs for the full 30s until the test harness's own watchdog force-closes it (a clean, bounded failure, not an unrecoverable hang).Plausible causes, not yet confirmed
onLoadingStateChange, which may be too early for the GL surface to be realized.e.getModifiersEx(),getClickCount()) not fully replicated by a bareComponent.dispatchEvent()the way a real OS-level event would carry them.onBeforePopup/issue JCEF doesn't expose CefExecuteJavaScriptWithUserGestureForTests, blocking real onBeforePopup test coverage #11, file-dialog trigger, download trigger) -- though those are specifically about user gesture requirements for navigation/download/dialog APIs, and it's unconfirmed whether context-menu triggering has an analogous requirement.Suggested next steps
CefRenderHandler.onPainthas fired at least once) rather than immediately on load completion.MouseEventfields a real OS-generated right-click carries (via manual/interactive testing) against what this synthetic one sets, to find a field-level mismatch.Found via
This fork's coverage-expansion effort (tracked in #5) --
CefContextMenuParams_N.cpp/CefMenuModel_N.cppare the single largest remaining 0%-covered chunk (481 lines combined) per a realgcovrmeasurement, and this was the natural next test to attempt.