Re: Suggestion for the VMS X-windows server


"JF Mezei" <jfmezei.spamnot@xxxxxxxxxxxxx> wrote in message
news:e3e25$45fda29a$cef8887a$22229@xxxxxxxxxxxxxxx
FredK wrote:
It isn't clear to me that it has "run out of resources". Becomming
unresponsive doesn't meet the criteria. If the X server "runs out of
memory" --- it crashes (or sometimes if you are lucky it kills a
connection) - it doesn't just "slow down".

With Mozilla, I see the X server's process raise its Working set to
wsextent, and raise its virtual pages to very near pgfilquota. Show
proc/quota reveals it has 256 pages left as pgfilquota.

At this point in time, mozilla does not redraw the window. Menus do not
appear, however, looking at the mozilla process (show proc/cont), I see
some tiny action when the mouse moves over a clickable area such as the
menu mar that should be just under the MWM window title. (MWM works fine
during those situations, it is just the content of all mozilla windows
that do not refresh).

Bringing other simpler windows to the foreground (such as decterm), they
work perfectly, and they redraw fine.

There are errors logged to decw$server_0_error.log about insufficient
resources (but I currently have no examples since I have not gone to any
complex/large web pages since restarting the x server).

I need an example.

Essentially, there was no real plan in X11 for dealing with running out of
resources like memory, and routine interfaces don't bubble status back up to
the callers (and ulitmately to a point where it could return an error to the
client like INSMEM). If it happens during a request at a point where the
code knows the client ID - then it can kill the client (very friendly). If
it happens in other places, it issues a fatal server call (even friendlier).

X11 assumes an error-free connection (no robustness features to allow a
re-connect) and error free execution of the X11 server with infinite
resources. It also assumes mostly error free clients that respond in a
timely manner to events to prevent deadlocks.





.

Relevant Pages

  • Re: Suggestion for the VMS X-windows server
    ... With Mozilla, I see the X server's process raise its Working set to wsextent, and raise its virtual pages to very near pgfilquota. ... Menus do not appear, however, looking at the mozilla process, I see some tiny action when the mouse moves over a clickable area such as the menu mar that should be just under the MWM window title. ... There are errors logged to decw$server_0_error.log about insufficient resources. ...
    (comp.os.vms)
  • Re: DECW$SERVER crashes (8.3)
    ... when a window is closed, does the X server automatically destroy all pixmaps associated with that window? ... Or are pixmaps self sustaining entities that can survive the destruction of one window so that the client can re-used it on another window (consider mosaic or mozilla which can have multiple windows opened at the same time). ... I had the server crash long after I had closed some image intejnsive page that couldn't display all images. ...
    (comp.os.vms)
  • Re: Suggestion for the VMS X-windows server
    ... The act of popping up an alert may well cause the failure. ... process/window has made the request that cause the X server to go below ... When mozilla goes nuts because it is trying to display some huge blog ... neither does the window manager's "close window". ...
    (comp.os.vms)
  • Re: TCP Window Size
    ... terminal server, so window size wouldn't be the problem. ... network I/O, active clients, etc. ... Multiple parallel WAN links are being load ...
    (comp.dcom.sys.cisco)
  • Re: Using Outlook with Lotus Notes R7
    ... provided Outlook Connector for IBMLotusDomino. ... unsuccesful in getting myLotusNotes email and calendar to display in ... On the Server Type page of the wizard, ... closed the Mail window. ...
    (microsoft.public.outlook.interop)