Re: Linux, X, ld, gcc, linking, shared libraries and stuff

På Thu, 31 Mar 2005 01:34:16 GMT, skrev Beth <BethStone21@xxxxxxxxxxxxxxxxxxxxxxx>:



The point about X is that it works exactly the same as Windows does
otherwise...but has the addition of an extra API call, where you can
specify an "event mask"...and then only those classes of events that you
specify are delivered to the application...

Why is this so hard for people to see and understand? It's very logical,
sensible and obvious...

hehehe, so _thats_ why many linux apps goes sort of dead... they dont get enough messages... The very thing that happend, to a basically empty Kylix app. It just went sort of dead. Refused to redraw itself, but could get resized... in limbo sort of..if i recall right.

It's really very simple and very logical...indeed, that is the only
criticism of Windows here, in a sense: Just how stupid Microsoft were to
miss such a simple, sensible "optimisation" there themselves...one extra
API call and a few extra "window attributes" to store that keep track of
what the "message filter" is set to...

Its not optimization. Linux just moved the filtering function (if one can rely on a single word from you) into the "os" instead. Now the Os just filters all msgs on behalf of an app. Hardly any more optimized, when you consider, that each app must register theese filters, maybe several times per app.

then Windows wouldn't deliver
it...indeed, it could _immediately_ send it to "DefWindowProc"...this would
then leave the application "blocked" and not woken up by any "false alarm"
messages...

This would be slower. As we know, even if windows is dreadful, it appear 3-4 times faster then linoliumX on the distros I have tested. Also I refer to the needless call to defwndproc, that could be omitted for the major bunch of messages. See my post in reply to Frank, for proof. When did, you ever post any proof ?


You see how this works yet? It's very simple and very logical and very
sensible, I feel...

What about do some coding to proove your feel. Lets see the messages posted to a Linux app when you move the mouse over it. Lets see a linux messageloop?


Do you remember my example ages back with "hotel reception" about blocking? 

with pain.

Where I'm waiting for a friend to turn up at the hotel? 

dream on :-)

Now, let's extend the example to include this new facet...I've told
reception that they should phone me when my friend arrives...so I go to
sleep ("blocked")...


truth is so revealing....



"Does
your friend have blonde hair?"
"Does your friend have blue eyes?"


And, then, every 10 seconds or so, the phone is ringing... 

You must have too many friends  ;-)

r...I'm
constantly being "nagged"


....but this, so to speak, is the Windows way of
doing it...

If you do the filtering in the os, or the _one_ app to receive the message, count for NOPE.
Are you going to tell Linux, theres a button here, and here, and here. what if one animate like my IQ test :-)), am I going to tell Linux about this change, for every move of the mouse, or button ? Sorry but I see no big win.
< http://szmyggenpv.com/downloads/IQ test.Zip >


This 
means that the system overall should only ever have processes "unblocked"
to do _actual useful work_...no "false positives" or "false alarms" at
all...

The filtering, must be done somewhere, and in linux you get to tell linux that the button has moved?? Yo did say, _exactly_ up there...

you can also modify the "mask"
dynamically...so, you know, you might have it that when a listbox is empty,
it "greys out" to show its disabled and then you can set its "event mask"
to just "paint messages" but ignore everything else...an entry is added to> the listbox, it turns to the white background and shows the entry...it's
enabled again...right, at the same time, you can now also add mouse clicks
and such to the "event mask"... 

Yeah we get the idea. More work for nope.

that is, basically, you could also - if 
you're the pedantic sort and can't bear to lose a single clock cycle to
anything "redundent"

But thats not who _you_ are, is it ? You are wasting bandwidth as well....:-))

)...although, if this all sounds like too much bother,

< http://szmyggenpv.com/downloads/IQ test.Zip >


Beth :)

P.S. Bloody vikings! 

Lol.



--
http://TheWannabee.org
.

Quantcast