WebKit Bugzilla
New
Browse
Log In
×
Sign in with GitHub
or
Remember my login
Create Account
·
Forgot Password
Forgotten password account recovery
NEW
264509
[WPE][GTK] API test /WPE/TestResources afterAll is a flaky crash
https://bugs.webkit.org/show_bug.cgi?id=264509
Summary
[WPE][GTK] API test /WPE/TestResources afterAll is a flaky crash
Carlos Alberto Lopez Perez
Reported
2023-11-09 10:45:05 PST
WPE failure: TEST: /app/webkit/WebKitBuild/WPE/Release/bin/TestWebKitAPI/WPE/TestResources... /webkit/WebKitWebView/resources: PASS GLib-GIO-DEBUG: Using cross-namespace EXTERNAL authentication (this will deadlock if server is GDBus < 2.73.3) EGLDisplay Initialization failed: EGL_NOT_INITIALIZED /webkit/WebKitWebView/history-cache: PASS /webkit/WebKitWebView/sync-request-on-max-conns: PASS /webkit/WebKitWebResource/loading: PASS /webkit/WebKitWebResource/response: PASS /webkit/WebKitWebResource/mime-type: PASS /webkit/WebKitWebResource/suggested-filename: PASS /webkit/WebKitWebResource/active-uri: PASS /webkit/WebKitWebResource/get-data: PASS /webkit/WebKitWebResource/get-data-error: PASS /webkit/WebKitWebResource/get-data-empty: PASS /webkit/WebKitWebPage/send-request: PASS (TestResources:3762): GLib-CRITICAL **: 14:23:35.177: g_atomic_ref_count_dec: assertion 'old_value > 0' failed GLib-FATAL-CRITICAL: g_atomic_ref_count_dec: assertion 'old_value > 0' failed [...] Unexpected crashes: /WPE/TestResources afterAll Log:
https://ews-build.webkit.org/#/builders/41/builds/2645/steps/26/logs/stdio
It happens randomly on the EWS queue mainly at:
https://ews-build.webkit.org/#/builders/API-Tests-WPE-EWS
Looks like a race condition somewhere.
Attachments
Add attachment
proposed patch, testcase, etc.
Carlos Alberto Lopez Perez
Comment 1
2023-11-09 10:58:41 PST
Funny that it crashes when trying to delete the object Tools/TestWebKitAPI/Tests/WebKitGLib/TestResources.cpp- Tools/TestWebKitAPI/Tests/WebKitGLib/TestResources.cpp:void afterAll() Tools/TestWebKitAPI/Tests/WebKitGLib/TestResources.cpp-{ Tools/TestWebKitAPI/Tests/WebKitGLib/TestResources.cpp- delete kServer; Tools/TestWebKitAPI/Tests/WebKitGLib/TestResources.cpp-} Seems a refcount or pointer issue somewhere. This happens very randomly, to reproduce it you have to run the tests in a loop until it eventually happens (1th of 100 times maybe)
Carlos Alberto Lopez Perez
Comment 2
2023-11-09 11:07:55 PST
This happens also on GTK bot. See:
https://ews-build.webkit.org/#/builders/21/builds/26577/steps/26/logs/stdio
TEST: /app/webkit/WebKitBuild/GTK/Release/bin/TestWebKitAPI/WebKitGTK/TestResources... /webkit/WebKitWebView/resources: PASS GLib-GIO-DEBUG: _g_io_module_get_default: Found default implementation local (GLocalVfs) for ‘gio-vfs’ GLib-GIO-DEBUG: Using cross-namespace EXTERNAL authentication (this will deadlock if server is GDBus < 2.73.3) /webkit/WebKitWebView/history-cache: PASS /webkit/WebKitWebView/sync-request-on-max-conns: FAIL ERROR:/app/webkit/Tools/TestWebKitAPI/Tests/WebKitGLib/TestResources.cpp:735:testWebViewSyncRequestOnMaxConns(SyncRequestOnMaxConnsTest*, gconstpointer)::<lambda(gpointer)>: code should not be reached /webkit/WebKitWebResource/loading: PASS GLib-GIO-DEBUG: _g_io_module_get_default: Found default implementation local (GLocalVfs) for ‘gio-vfs’ GLib-GIO-DEBUG: Using cross-namespace EXTERNAL authentication (this will deadlock if server is GDBus < 2.73.3) /webkit/WebKitWebResource/response: PASS /webkit/WebKitWebResource/mime-type: PASS /webkit/WebKitWebResource/suggested-filename: PASS /webkit/WebKitWebResource/active-uri: PASS /webkit/WebKitWebResource/get-data: PASS /webkit/WebKitWebResource/get-data-error: PASS /webkit/WebKitWebResource/get-data-empty: PASS /webkit/WebKitWebPage/send-request: PASS (TestResources:23087): GLib-GIO-CRITICAL **: 19:24:54.755: g_socket_accept: assertion 'G_IS_SOCKET (socket)' failed GLib-GIO-FATAL-CRITICAL: g_socket_accept: assertion 'G_IS_SOCKET (socket)' failed
Carlos Alberto Lopez Perez
Comment 3
2023-11-09 11:49:07 PST
Updated expectations on
270468@main
Note
You need to
log in
before you can comment on or make changes to this bug.
Top of Page
Format For Printing
XML
Clone This Bug