Re: Realtime signals in glibc version 2.3.1

[Andreas Jaeger]

address@hidden writes:

> From: Andreas Jaeger <address@hidden>
>> 
>> Stephen Rothwell <address@hidden> writes:
>> 
>> > The following program used to print 32 and 63 on ia32 Linux with
>> > glibc version 2.2.5 (unless linked with -lpthread when 35 and 63
>> > were printed).  With version 2.3.1, -1 and -1 are printed.  This
>> 
>> You need to link against libpthread if you like to use librt,
>
> I don't want ot use librt or libpthread, I just want to use the
> kernel's ability to queue signals above SIGRTMIN.  This used
> to work with no fuss.  Was this an arbitrary decision, or is there
> some standard you are following here that we used to break?

Now I see what you mean - and don't have directly any help for you.
>
>> > Also, if <unistd.h> happens to be included, the _POSIX_REALTIME_SIGNALS
>> > is defined, but SIGRTMIN remains -1 (unless linked against -lpthread).
>
> My point here is that if you are trying to follow the relevant
> Posix standard, then including <unistd.h> will give the programmer
> the impression that realtime signals are supported, but SIGRTMIN
> is stiil defined as -1 unless -lpthread is used ...

please file a complete bug using the glibcbug script so that it's in
our bug database,

Andreas
-- 
 Andreas Jaeger
  SuSE Labs address@hidden
   private address@hidden
    http://www.suse.de/~aj