Building without thread support

Peter Kjellerstedt peter.kjellerstedt at axis.com
Mon Sep 25 07:44:21 UTC 2006


> -----Original Message-----
> From: uclibc-bounces at uclibc.org 
> [mailto:uclibc-bounces at uclibc.org] On Behalf Of Jim Blandy
> Sent: Saturday, September 23, 2006 06:07
> To: Peter Kjellerstedt
> Cc: uclibc at uclibc.org
> Subject: Re: Building without thread support
> 
> Mike Frysinger <vapier at gentoo.org> writes:
> > On Friday 22 September 2006 12:51, Peter Kjellerstedt wrote:
> >> While I was looking for a another problem, I tried to build
> >> the latest svn version of uClibc without thread support.
> >> This failed on the first file it tried to compile, and I had
> >> to apply the following patch to make it compile.
> >> Any comments, or should I go ahead and commit it?
> >
> > certainly ... we should even take it another step and wrap 
> > it in a TLS check as not all architectures have TLS support yet ...
> 
> While you're at it, Peter --- there are actually *two* inclusions of
> <tls.h> in that file; you may want to factor them out into one
> #inclusion.

Sounds like a good idea. However, I do not think I am the right 
person to do it, as I do not know this part of the code, and do 
not know what is the right thing to do.

> That said, if I understood right, Peter Mazinger's plan is to make a
> new header, include/tls.h, that can be safely #included by any code in
> uClibc, regardless of what thread model the user has chosen, if any.
> It'll then define the Right Things to let you know whether you can use
> __thread.  So following that plan, the fix would be to move the checks
> into a new include/tls.h.

Unfortunately I believe PSM has abandoned all work on (the official)
uClibc.

> I'm guessing the mistake that broke the build was that code assuming
> an all-model <tls.h> got placed in uClibc_errno.h before the actual
> <tls.h> itself was added.  :)

//Peter



More information about the uClibc mailing list