Re: Is buildworld broken?

[Chris Pressey <cpressey@xxxxxxxxxxxxxxx>]

On Sun, 20 Mar 2005 02:27:51 +0900
YONETANI Tomokazu <qhwt+dfly@xxxxxxxxxx> wrote:

> On Sat, Mar 19, 2005 at 11:04:26AM -0500, Adrian Nida wrote:
> > I'm trying to rebuild world, kernel, etc. and ran into this error.
> > I rm'ed -r / usr/obj, rm'ed -r /usr/src, and re-cvsupped (as of ~30
> > minutes ago).  Is anyone else seeing this?  If not, any ideas what
> > I should do to fix?
> 
> > /usr/lib/libc.a(getopt_long.o): In function `getopt_long':
> > getopt_long.o(.text+0x86c): multiple definition of `getopt_long'
> > getopt_long.o(.text+0x864): first defined here
> 
> The libc currently installed on your system contains getopt*
> functions. This probably means that the last world installed was built
> from source between "2005-03-13 20:13:41" and "2005-03-14
> 14:26:16"(see joerg's commit on commits@).

Joerg: this is what I was talking about in the "Buildworld error"
thread.

Adrian: Updating my source tree didn't work for me when I ran into this
- I had to fix this by downloading a snapshot ISO, mounting it with
vnconfig, and copying /mnt/usr/lib/libc.* off of it into /usr/lib.

-Chris

> Try checking out the source slightly after "2005.03.14 14:27:00(UTC)"
> and build and install libc, then start buildworld of -CURRENT.
> 
> This reminds us that, when our getopt() is replaced with
> getopt_long*(), we need a way to know if the building system has
> getopt_long* family in its libc, or buildworld on months older system
> or on FreeBSD 4.x fails at the bootstrapping stage.
>