BIND 9.10.1rc2 won't build on FreeBSD 10-STABLE
john.marshall at riverwillow.com.au
Fri Sep 12 09:52:15 UTC 2014
On Thu, 11 Sep 2014, 23:38 +0000, Evan Hunt wrote:
> On Fri, Sep 12, 2014 at 09:11:08AM +1000, John Marshall wrote:
> > I can't build BIND 9.10.1rc2 on recent FreeBSD 10-STABLE.
> > I have tried on both i386 and amd64 variants of the operating system.
> > BIND 9.10.1rc1 builds fine, as did the beta releases.
> Based on the failure being in bin/python, I suppose it was this:
> 3946. [cleanup] Improved "configure" search for a python interpreter.
> [RT #36992]
> I'm guessing your system doesn't have a python interpreter, but configure
> got confused into thinking it does. If I'm right, then installing python
> ought to make the build work, for the time being. We'll address the
> problem before final release.
Sorry for the brief report earlier. I had to run out the door to an
appointment and didn't have time to provide more information but wanted
at least to send a heads-up.
I did some more digging. Ironically, it seems that the improved
configure search is finding python in rc2 whereas the rc1 and earlier
configure did not find python. I hadn't tested these new utilities at
all so I was unaware that they were not being built in the earlier rc or
betas. It still looks to me like a make(1) problem but it only kicks in
when python is detected and an attempt is made to build in bin/python.
A FreeBSD 9.3-RELEASE system is building rc2 happily: it uses pmake: but
bmake on the 10-STABLE systems falls over with:
making all in /build/bind/bind-9.10.1rc2/bin/python
make: don't know how to make dnssec-checkds. Stop
make(1) on the 10-STABLE systems cannot find the source for the
dnssec-checkds target. Comparing the make(1) debug output between the
two systems, it appears that pmake on the FreeBSD 9.3 system picks up
the .py transform as a clue for a source for a null-suffix target, but
bmake on the FreeBSD 10-STABLE systems does not include .py in the list
of candidate source suffixes for null-suffix targets.
Sorry, I'm not a make(1) guru and can't suggest a solution apart from
putting explicit dependencies in the Makefile. I'm sure there must be a
more elegant solution which coaxes bmake to behave as expected.
> Do you still have your config.log? May I see it?
Sure. I've included copies of the complete build directories
(post-make) from both rc1 and rc2 and made them browsable.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 181 bytes
Desc: not available
More information about the bind-users