Re: DECW$SM fills up the disk---is this a known bug?
From: Martin Kirby (martinkirby12_at_yahoo.co.uk)
Date: 08/20/03
- Next message: _at_SendSpamHere.ORG: "Re: AntiVirus for OpenVMS"
- Previous message: Christoph Gartmann: "Re: Strange programming problem"
- In reply to: Phillip Helbig: "Re: DECW$SM fills up the disk---is this a known bug?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 20 Aug 2003 03:48:47 -0700
Phillip,
I can try and explain in more detail but I'm not sure how much success
I'll have :-).
When the session manager starts the screen saver it creates a window
that covers the whole screen and passes the window id of that to the
screen saver process. This is also how the style manager can run the
screen saver in a small example window. You can see this window if you
run DECW$UTILS:XLSWINS when the screen saver is displayed (having
logged in remotely). It has a name something like "Cover Dialog".
The screen saver does its screen output, bouncing lines, fireflies or
just a black area, to this window. If the window doesn't exist then it
gets repeated X window errors and just keeps trying to write to the
window (the fix is to exit after an error). The X errors are output to
the log file.
When the screen saver starts it creates an invisible window and sends
the identity of that to the session manager by sending an event to the
cover window. The session manager records the screen savers window id
and when the session manager unlocks it uses this to send an
XKillClient request, prior to deleting the cover window.
One race condition is that the session manager processing is:
1. Start processing event that caused unlock
2. Kill registered screen saver clients (multiple
if have multi-screen)
3. Delete cover windows
If the screen-saver was just started and its processing sends the
event between (1) and (3) then the session manager never sees the
event because the window was deleted [event processing uses Xt and
that ignores events on windows the widgets of which have been deleted]
and the screen saver process is not terminated.
There are other race conditions, some of which are allowed for in the
code and some of which are not.
Hope that is some help,
Martin Kirby
DECwindows Engineering
- Next message: _at_SendSpamHere.ORG: "Re: AntiVirus for OpenVMS"
- Previous message: Christoph Gartmann: "Re: Strange programming problem"
- In reply to: Phillip Helbig: "Re: DECW$SM fills up the disk---is this a known bug?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|