Path: utzoo!attcan!uunet!mcvax!guido
From: guido@cwi.nl (Guido van Rossum)
Newsgroups: comp.windows.x
Subject: Re: twm reactions
Message-ID: <371@piring.cwi.nl>
Date: 15 Jun 88 09:07:06 GMT
References: <8806141942.AA29931@ATHENA.MIT.EDU> <10952@agate.BERKELEY.EDU>
Reply-To: guido@cwi.nl (Guido van Rossum)
Organization: The Royal Society for Prevention of Cruelty to Amoebae
Lines: 27

In article <10952@agate.BERKELEY.EDU> rusty@velveeta.UUCP (rusty wright) writes:
>>First, the pulldown menus are not as quick as desirable (on my sun 3).
>[...]
>I'm pretty sure that the reason it's so slow to bring up the menu the
>first time when I haven't used it for a while is because I'm using a
>sun 3/50 (only 4 megs of memory) and twm has simply either been swapped
>out or most of its pages have been paged out and the delay is because
>unix is paging or swapping it in.

It may also be that the server is taking some time to map in umpteen
windows and send them all exposure events.  I believe one window takes
0.5K or 1K bytes in the server; a TWM menu with ten items would thus
take 6 or 12K.  Multiply this with the number of menus you have
(assuming TWM just maps/unmaps the window) and you get quite a bit of
memory devoted to menus in the server, which may also be paged out.
I suppose the 12 exposure events (and possibly also 12 map events?) also
take their toll.

I have written a toolkit with pull-down menus which takes a different
approach: the menu is a single window and I track button motion events
to detect when the pointer enters/leaves an item.  Highlighting feedback
is not slower than with other strategies, and the menus pop up much
faster, even though I never keep any unmapped windows; I just create and
destroy the menu's window as needed.  Anyone care to comment?
--
Guido van Rossum, Centre for Mathematics and Computer Science (CWI), Amsterdam
guido@piring.cwi.nl or mcvax!piring!guido or guido%piring.cwi.nl@uunet.uu.net