Seeing this in a VMWare guest, too.  I think in add_session_record(), this check should be "||" instead of "&&"


if (display_name == NULL && display_device == NULL) {
    goto out;
}

(In reply to Sinclair Yeh from comment #1)
> Seeing this in a VMWare guest, too.  I think in add_session_record(), this
> check should be "||" instead of "&&"
> 
> 
> if (display_name == NULL && display_device == NULL) {
>     goto out;
> }

Actually, strike that.  The issue is gdm_session_record_logout() is not checking x11_display_name before using it.  Perhaps there should be a check like the one at the beginning of gdm_session_record_login()

yea, totally.  I goofed on that one.

Created attachment 306483 [details] [review]
session-record: support NULL display name if tty available (for logout/failed)

wayland sessions don't necessarily have a display name, so this
commit just uses the display device instead.

It's like commit 1edb0e4ae897cb1c53af3f3b2b913d5baf731580 but for
logout and failed records.

Discovered by Sinclair Yeh <syeh@vmware.com>

Attachment 306483 [details] pushed as 29ecdb3 - session-record: support NULL display name if tty available (for logout/failed)

The patch looks good to me.

Reviewed-by: Sinclair Yeh <syeh@vmware.com>

This looks like a good candidate for a backport?

indeed, i'm overdue for a 3.16 release anyway, doing one now