Hello Debian GNOME maintainers, Follow-up build evidence for my previous source-level RFC sent to this bug.
I compiled gnome-keyring 51.1 with a guarded patch that replaces a NULL-PKCS#11-session g_return_val_if_fail() in the two OpenSession negotiation paths (AES and plain) with g_set_error_literal(error_out, G_DBUS_ERROR, G_DBUS_ERROR_FAILED, ...) followed by return FALSE. The source archive was verified: SHA256 dfe7adc5876ef0670e7eba2907799bdbb8875ce58bb4208e1a52169bd16a47ca Build was isolated inside an Ubuntu 26.10 Docker container; the host's keyring and credentials were not mounted. Meson 1.10.1 configuration succeeded, ninja compilation passed, and a gnome-keyring-daemon binary was produced. Test summary: 60/61 Meson tests passed. The only failing test was gnome-keyring:startup, /daemon/startup/control/noaccess. It checks that an inaccessible control directory is not accepted, but in the root-run container the test process could access the directory and its assertion failed (the expected and actual control paths were identical). Root execution is a plausible environment cause; it has NOT been independently verified because the non-root rerun was blocked. This is not an assertion of full test-suite PASS. Build log excerpt: 60/61 gkr-daemon - gnome-keyring:startup FAIL SIGABRT assertion failed (g_environ_getenv(output, "GNOME_KEYRING_CONTROL") != directory) Summary: Ok: 60, Fail: 1 A synthetic missing-client D-Bus reproducer and an end-to-end verification of the crash fix are still missing. I did not install the compiled binary on the affected machine. Please compare with GNOME MR !112 (OpenSession client-record handling) before treating this as a separate patch. Regards, Alexander Stepanov
