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

List:       kde-core-devel
Subject:    Re: threads
From:       Roberto Alsina <ralsina () unl ! edu ! ar>
Date:       1999-12-02 11:12:21
[Download RAW message or body]

On Thu, 2 Dec 1999, Lubos Lunak wrote:

> Well, since now there is GNU Pth, we can have threads for all KDE platform. And
> if KDE has to run on all these platforms, we'll have to write the apps as if
> the only underlying Pthreads library was Pth, since Pth is cooperative and
> thus limited compared to let's say LinuxThreads. I mean, code which assumes
> preemptive threads scheduling will block with Pth ( although using threads just
> for better responsive GUI will behave with Pth the same way it would without
> threads - it would just block for a white ).

Well, not all the time if you use one of pth's system call overriding
modes (I don;t think that's actually how they are called), that turns
calls to read() into calls to pth_read() and such.

That way the only real blocks would be in tight loops with no yield()s
inside of them, and those are easy to spot and fix.

> And if will have to assume Pthreads lib is actually Pth, why not use Pth for
> all KDE platforms ? This way facts like Qt not being thread-safe ( btw, when it
> will be finally MT-safe ? ) or X11 libs not being thread-safe are irrelevant -
> they will be ok with Pth. It would probably block more often than with
> LinuxThreads, but this way there would be less problems. Also, since Pth is
> cooperative, writing MT apps would be much easier, debugging would be easier, 
> and one would produce less bugs ( or they wouldn't have chance to break anything
> ).

YES. Trying to figure out all race conditions on a "real" multithreading
program is a pain.

For example, you can't have any globals at all, or you are begging for
problems, and, while a little ugly, I find globals amazingly useful :-)

 ("\''/").__..-''"`-. .         Roberto Alsina
 `9_ 9  )   `-. (    ).`-._.`)  ralsina@unl.edu.ar
 (_Y_.)' ._   ) `._`.  " -.-'   Centro de Telematica
  _..`-'_..-_/ /-'_.'           Universidad Nacional del Litoral
(l)-'' ((i).' ((!.'             Santa Fe - Argentina
                                KDE Developer (MFCH)
Not mad, but bound more than a madman is (Romeo and Juliet, Act I Scene II)

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