GNU uname should identify OS as illumos

[jclulow]

As of bug 1956, the uname binary provided by the operating system will report illumos for uname -o instead of the traditional Solaris. Unfortunately in a pkgsrc-based zone, at least under some conditions the GNU uname appears first in PATH. A naive script may run uname -o and see the Solaris value.

We should patch the pkgsrc copy of uname to report illumos as the system version does.

See example with current trunk packages:

$ /usr/bin/uname -o
illumos
$ /opt/local/bin/uname -o
Solaris

[wiedi]

A long time ago I played with this and had the following:

diff --git a/sysutils/coreutils/Makefile b/sysutils/coreutils/Makefile
index 82c440e..cdf8ce7 100644
--- a/sysutils/coreutils/Makefile
+++ b/sysutils/coreutils/Makefile
@@ -43,6 +43,11 @@ SUBST_FILES.libname= Makefile.in
 SUBST_SED.libname=     -e '/^src_libstdbuf_so_LDFLAGS/s,$$, -install_name ${PREFIX}/libexec/coreutils/glibstdbuf.so,'
 .endif
 
+.if ${OPSYS} == "SunOS"
+CONFIGURE_ENV+=                gl_cv_host_operating_system=illumos
+.endif
+
+
 PROGS= [ base64 basename cat chcon chgrp chmod chown chroot cksum comm \
        cp csplit cut date dd df dir dircolors dirname du hostid echo env \
        expand expr factor false fmt fold groups head id install join kill \

But that is lazy way to fix it as it would be wrong when using pkgsrc on Oracle Solaris.
Propably should just try to shell out to /usr/bin/uname -o now?

[jperkin]

Older Solaris releases don't support -o so we'd need to handle that, but yeh that seems the simplest way to do it.

[jperkin]

Note that I will need to override this for our builds anyway as they are performed on a platform prior to 1956.

[jperkin]

Fixed in trunk.