Re: does libfetch need pthreads?

["Joachim Kuebart" <joachim.kuebart%gmx.net@localhost>]

-------- Original-Nachricht --------
> Datum: Thu, 5 Mar 2009 16:15:34 +0100
> Von: Joerg Sonnenberger <joerg%britannica.bec.de@localhost>
> An: pkgsrc-users%netbsd.org@localhost
> Betreff: Re: does libfetch need pthreads?

> On Thu, Mar 05, 2009 at 04:06:01PM +0100, Joachim Kuebart wrote:
> > It would appear they're only hiding the prototype in <time.h> as the
> > symbols do appear in libc. At any rate, adding -D_REENTRANT was all I
> > needed to do to make it build successfully.
> 
> OK, I've added the define.

Thanks, that fixes it.

> > Having said that, bootstrapping on Interix currently fails anyway
> > because of an antique <sys/queue.h> which lacks TAILQ_FIRST and
> > TAILQ_LAST and has broken block macros like TAILQ_INSERT_TAIL
> > (without "do {...} while(0)") which break when used inside if/else,
> > for example, in pkg_delete.c. Not sure what can be done about that,
> > personally I simply replaced <sys/queue.h> with a less broken version.
> 
> As long as the macros are missing, nbcompat should provide them. The
> issue with TAILQ_INSERT_TAIL has been fixed already, Linux has similar
> broken macros. If you have a current tree and it still happens, please
> provide the full error.

You're right, the most recent version doesn't have the problem anymore,
sorry about the noise. My snapshot was a few days old as I've only just
finished installing CVS :-).

Confirming this took me a while before I figured out that I needed to set
UNPRIVILEGED_GROUP to my numeric group id in order for testbootstrap to
succeed. This is because the default "id -gn" from mk/unprivileged.mk
comes up with "Domain Users" on my Interix which contains a space and all
hell breaks loose. On Interix, it seems necessary to use "id -g" instead.

Best regards,

Joachim