sigprocmask testing program failed with static linking

Rich Felker dalias at aerifal.cx
Sat Apr 23 21:38:05 UTC 2011


On Sat, Apr 23, 2011 at 01:25:59PM +0200, Peter Mazinger wrote:
> I am all for removing duplicated functions in libc/libpthread, began
> this in future branch (cancellable functions), if we need separate
> implementations, I would prefer to rename functions in libpthread to
> "show us", that this is something else...

Are they "duplicated"? The traditional reason these functions exist is
that either POSIX or the nature of the pthreads implementation imposes
additional requirements on the function. The "plain" function then
gets linked in non-threaded applications, and the "pthread" version
gets linked if -lpthread is used. If you remove the pthread version,
you WILL BREAK THINGS BADLY. Removing the "plain" version should not
break anything except for pulling in possible dependencies on other
pthread functions.

I hope eventually uClibc can drop the whole practice of treating
threads as an "add-on" and instead integrate the code better (in a way
that doesn't pull in unnecessary deps) but so that only one
implementation is needed.

Rich


More information about the uClibc mailing list