[vdr] Re: [FINAL TESTING]: text2skin 1.0-rc1

["Gunnar Roth" <gunnar.roth@gmx.de>]

----- Original Message ----- 
From: "Klaus Schmidinger" <Klaus.Schmidinger@cadsoft.de>
To: <vdr@linuxtv.org>
Sent: Monday, January 03, 2005 3:40 PM
Subject: [vdr] Re: [FINAL TESTING]: text2skin 1.0-rc1


> Gunnar Roth wrote:
> > ...
> > > Rainer Zocholl wrote:
> > >
> > > > Klaus.Schmidinger@cadsoft.de(Klaus Schmidinger)  03.01.05 10:51
> > > >
> > > >
> > > > > ...
> > > > > Just for the record: there will be _no_ poisoning in the official VDR 
> > > > > source.
> > > >
> > > >
> > > > Why?
> > >
> > > Because VDR will no longer use any non-threadsafe library functions!
> > >
> > > > Do you like to hunt phantom errors and like unreliability? ;-)
> > > >
> > > > I don't see no reason to place such mean booby traps for the
> > > > other programmers in the code/make file.
> > > >
> > > > So what's the reason to allow the usage of non threadsafe non
> > > > reentrant libray call in a heavily multi threaded environment?
> > >
> > > Since VDR won't use any non threadsafe library functions, what
> > > would be the point in "poisoning"?
> > to "guarantee" that none of the non threadsafe functions are used maybe?
> > to force plugins not to use this non-thread functions?
> >
> > regards,
> > gunnar
> So whenever there's a new non-threadsafe library function, would you
> expect VDR to "poison" that one, too? Even though by itself it doesn't
> use any of them?
1. What new library functions? I dont think that this kind of 
functions(being not threadsafe)
will be added to glibc.the existing functions are left from ancient history.
(btw. does anyone know why glibc does not use thread local storage 
internally (as ms crt does) to make this kind of functions
threadsafe, but uses xx_r named alternatives?)
2. If vdr does not use one of the functions yet, it may be used in the 
future, so it easily reminds you of not using this function.

> Sorry, but that makes no sense to me.
For me it does.

> If plugin authors want to avoid non-threadsafe functions (which they 
> should,
> of course), they need to be careful by themselves. Maybe there should be
> some compiler switch that flags such functions as errors, but don't put
> the burdon on VDR.
Why is it a burden? Its already done by others.
It helps plugin authors who come from other os, for example.
and it helps vdr, because it is at least guaranteed that a plugin does not 
use
unsafe functions. ( i know you only are responsible for vanilla vdr, but in 
users  perception this may be not the case.)

gunnar