[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: kmapnotify, LD_PRELOAD, (t)csh and aaarrgh! I'll die today! :-)
From:       tibirna () kde ! org
Date:       2000-05-29 2:59:23
[Download RAW message or body]


Thanks, Michael, for the good detailing. I have good proof that I was well
understanding the mechanism. It was just the point about the dlsyms'
precedence that I was overlooking.


On Mon, 29 May 2000, Michael Matz wrote:

> * we need to know, when the first window of an app apears
> - an window of an X-app appears, if either XMapWindow() or XMapRaised() is
>   called within that app (they are functions in libX11.so)

This automatically excludes non-X apps. Still, I'm sure that right now
notification attempts are done even when writing `bash` in the minicli
(without checking on 'Terminal' in options). I know that in this case the
notification times out after a while, but it's very plausible that in the
meanwhile the user used and even closed that non-X app :-) Is it possible
to detect if the app wants libX11 dlsyms?

[cut excellent explanation]

> Note that this works for _every_ application which uses X11 functions to
> map a window. We only need the binary for it. We don't need to patch any
> source, and no binaries. Note also, that this is the _only_ way that works
> for every app due to the way the dynamic linker works.

You mean the LD_PRELOAD way? I didn't finish my investigations. Isn't
there any ldl API that does exactly the same *in the code*, without
needing a SHELL for it?

> Hope this answers most of the questions ;)

The remaining ones (and recapitulation) are:  :-)

- is there a way to detect that an app isn't linked to X11 (`ldd` in the
API) so that kmapnotify doesn't start notification for nothing?

- is there a way to implement functionality similar to LD_PRELOAD directly
in the code?

- is it ok to think about implementing the kmapnotfy function directly in
KRun?

Thanks a lot for your incredible patience, Rik, Waldo, Michael.

Cristian Tibirna

[prev in list] [next in list] [prev in thread] [next in thread]