This is the mail archive of the
libc-hacker@cygnus.com
mailing list for the glibc project.
Re: ["Carlos Wijders" <computec@sr.net>] libc/968: 'make check' fails with glibc-2.1
- To: Andreas Jaeger <aj@arthur.rhein-neckar.de>
- Subject: Re: ["Carlos Wijders" <computec@sr.net>] libc/968: 'make check' fails with glibc-2.1
- From: Zack Weinberg <zack@rabi.columbia.edu>
- Date: Wed, 17 Feb 1999 11:24:10 -0500
- cc: libc-hacker@cygnus.com
On 17 Feb 1999 15:51:33 +0100, Andreas Jaeger wrote:
>>>>>> Andreas Schwab writes:
>
>AS> Andreas Jaeger <aj@arthur.rhein-neckar.de> writes:
>AS> |> Therefore we shouldn't use $(native-compile) for isomac (and neither
>AS> |> for posix/annexc).
>
>AS> This is wrong. Both are designed to be run on the *host*, not the target.
>AS> That means that they can even run when cross-compiling. The sources of
>AS> the programs are independent of the target environment. They only run the
>AS> target compiler as a child.
>
>The problem is not running those binaries but compiling them!
>
>The glibc2 FAQ advises to move /usr/include away when upgrading from
>libc5 to glibc2 - and if you do so, make check will fail since those
>two binaries can't be *compiled*. If they could be compiled, there
>wouldn't be a problem.
This mess is fixed in my revised test framework. The header tests are
compiled without reference to the system headers, and examine only the
headers provided by the just-built libc plus the compiler's headers.
zw