Shawn Walker wrote:
> FYI.
>
>
> ---------- Forwarded message ----------
> From: Aubrey Li <aubrey at blastwave.org>
> Date: Fri, May 9, 2008 at 3:38 AM
> Subject: [indiana-discuss] OpenSolaris 200805 build ON passed
> To: indiana-discuss Discuss <indiana-discuss at opensolaris.org>, Mark
> Haywood <Mark.Haywood at sun.com>, Rafael Vanoni <Rafael.Vanoni at sun.com>
>
>
> As I reported several days ago, OpenSolaris RC2a can't build ON.
> The issue exists in the official release as well. Now the root cause
> was found, and OpenSolaris 200805 can build ON successfully!
>
> As Mark pointed out, libdtracestubs.so isn't being created correctly.
> ==============================================
> nm -u ./obj64/unix.o ../../intel/genunix/obj64/libgenunix.so | grep
> __dtrace_pro
> be_ | sort | uniq | nawk '{ \
> printf("\t.globl %s\n\t.type %s, at function\n%s:\n", \
> $1, $1, $1); }' > obj64/dtracestubs.s
> ==============================================
> The problem is that dtracestubs.s isn't being created correctly.
>
> Here, nm is the root cause.
>
> When we build ON on SXCE,
> #which nm
> /usr/ccs/bin/nm
>
> And on Opensolaris 200805,
> #which nm
> /usr/gnu/bin/nm
>
> The really problem is, these two nm have different output.
>
> So, the workaround is, before build the ON, execute
> "export PATH=/usr/ccs/bin:$PATH"
>
> Now ON get the right "nm" and can be built properly.
Ah.. which explains why I had no problems, since my path doesn't have
gnu first.
This is a bug in the Makefile of course; it should read $(NM) rather than
nm. $(NM) is set in usr/src/Makefile.master:
NM= /usr/ccs/bin/nm
- Bart
--
Bart Smaalders Solaris Kernel Performance
barts at cyber.eng.sun.com http://blogs.sun.com/barts
"You will contribute more with mercurial than with thunderbird."
