[bind10-dev] Boost has to be upgraded on some build machines

[Jeremy C. Reed]

On Sat, 17 Aug 2013, Mukund Sivaraman wrote:

> On Fri, Aug 16, 2013 at 03:38:22PM -0500, Jeremy C. Reed wrote:

I was confused a little in preparing my responses as several lines were 
attributed to me but I never sent them.  I removed some of this from my 
reply.

> > http://git.bind10.isc.org/~tester/builder/BIND10/20130816202341-Solaris10-sparc-GCC/logs/files/config.log  
> > (see the configure:16904: checking Boost managed_mapped_file compiles)
> 
> Isn't Solaris the one platform where we configure --without-shared-memory?

On Solaris we have these combinations:

Solaris 10 on sparc using GCC

Solaris 10 on sparc using Sunstudio and using --without-shared-memory

Solaris 11 on i386 using GCC

> > http://git.bind10.isc.org/~tester/builder/BIND10/20130816200230-Debian6Linux-i686-GCC/logs/build.out
> 
> This is due to <https://svn.boost.org/trac/boost/ticket/7990> that
> apparently didn't make it into Boost's master branch (and 1.54.0). Would
> dropping to 1.53 help? Or you'd have to find the Boost patch and apply
> it locally.

Okay I patched it. Thanks.

> 
> > http://git.bind10.isc.org/~tester/builder/BIND10-distcheck/20130816195001-FreeBSD8-i386/logs/distcheck.out

This was quoted as coming from my sent email, but I didn't send it.  I 
removed this unrelated details.

> > http://git.bind10.isc.org/~tester/builder/BIND10/20130816193450-Solaris10-sparc-Sunstudio/logs/build.out
> >
> > "../../../../src/lib/log/log_formatter.h", line 198: Error:
> > isc::dns::AbstractRRset::AbstractRRset(const isc::dns::AbstractRRset&)
> > is not accessible from
> > isc::log::Formatter<isc::log::Logger>::arg<isc::dns::AbstractRRset>(const
> > isc::dns::AbstractRRset&).
> > "../../../../src/lib/log/log_formatter.h", line 198: Error: Formal
> > argument arg of type isc::dns::AbstractRRset in call to
> > boost::lexical_cast<std::string,
> > isc::dns::AbstractRRset>(isc::dns::AbstractRRset) has an inaccessible
> > copy constructor.
> 
> What version of Boost is used here?

It is 1.53.0. (The previous comment was lost above in the reply. It is: 
On the sol-10 (sparc) builder system, I upgraded from 
boost-headers-1.46.1 to boost-headers-1.53.0nb1 (using pkgsrc plus 
commented out all the configure settings in the installed user.hpp). 
...)

> (On that topic, we should get configure to produce the versions of
> detected dependencies in config.log output.)

Yes that would be good. I thought I had a ticket for it, but can't find 
it. I now created one based on my previous notes. 
http://bind10.isc.org/ticket/3107

> > http://git.bind10.isc.org/~tester/builder//BIND10/20130816164444-Debian6Linux-i686-GCC/logs/build.out

This also was not in my sent email (but I had linked to a newer output).

> This again seems to be a Boost issue. Can you upgrade to the latest
> version of Boost (1.54 if it works with the patch for boost ticket #7990
> applied - see above, or 1.53) ?

Per my sent email, I tried 1.52.0 and 1.54.0. I patched it now (see 
above).

> > http://git.bind10.isc.org/~tester/builder/BIND10/20130816180202-Solaris10-sparc-GCC/logs/configure.out

This also wasn't in my sent-mail. I had sent a link for a later failure.

> I am guessing configure is unable to find Boost here.
I may have overlooked this and didn't realize what happened, but I 
assume this happened at the moment when I switched between versions. 
Sorry I should have disabled run at that moment. I don't see same 
problem in the later runs.

> > http://git.bind10.isc.org/~tester/builder/BIND10/20130816175129-OpenBSD5-amd64/logs/configure.out

This also wasn't in my sent email.

> If OpenBSD doesn't provide a new Boost, it would be better to just point
> at an extracted tree of the latest version (unless there are any other
> issues that occur using this approach).

I extracted 1.54.0 and patched for their ticket 7990 and am now using it 
and removed the without-shared.

> > http://git.bind10.isc.org/~tester/builder/BIND10/20130816174502-Solaris10-sparc-Sunstudio/logs/build.out
> 
> There is no compiler output about what caused this. Did the compiler crash?

I don't know. The later failures are different and add more details.

(I removed the rest.)

  Jeremy C. Reed
  ISC