RWS@ZERMATT.LCS.MIT.EDU (Robert Scheifler) (10/26/87)
Date: Sun, 25 Oct 87 18:26:33 PST
From: deboor%bay.Berkeley.EDU@berkeley.edu (Adam R de Boor)
SYNOPSIS:
Frozen devices are supposed to remain frozen beyond the life of the
grab that froze them (at least, that's what I get from the definition
of GrabPointer in the protocol spec: "If keyboard-mode is Synchronous,
the keyboard (as seen via the protocol) appears to freeze, and no
further keyboard events are generated by the server until the grabbing
client issues a releasing AllowEvents request.") This is manifestly
not implemented in events.c.
REPEAT-BY:
Grab a button with keyboard_mode GrabModeSync, suspend the client,
press the button and release it, then type. The keystrokes will
continue to be transmitted, in spite of the fact that the keyboard
is supposed to remain frozen until the grabbing client has explicitly
issued an XAllowEvents request.
The intent of the protocol from the start was that the release of a grab
also undid any freeze it had caused on the other device. Early versions
tried to describe this explicitly in terms of AllowEvents requests, but
that changed with time. Apparently something got lost in the
translation. I believe the implementation is correct (I was the last
one to rewrite those routines, and I tested them a fair amount). I will
assume, unless I hear otherwise from significant people, that what needs
fixing is the wording in the spec.