[bind10-dev] BIND 10 trac3160, updated. 6faa17ef4f22d11d3fc653680d073ac61197d6b3 [3160] Log text representation of the AbstractRRset (contd.)

[Mukund Sivaraman]

Hi Michal

On Mon, Sep 16, 2013 at 09:17:49AM +0200, Michal 'vorner' Vaner wrote:
> Why is this change needed? Shouldn't .arg use the generic << operator? Shouldn't
> rrsets print themselves with that? I think that was the original intention and
> if it doesn't work, it is broken and we should fix that instead of working
> around it.

You know how we are facing issues with compilers and Boost. This work is
currently in a branch.

These changes are for issues with the latest version of stock Boost
(according to Jeremy) and SunStudio compiler. boost::lexical_cast seems
to be horribly broken on it:

1. It insists on making a non-const copy object of the object being
casted. It starts looking for the copy-constructor and fails because the
copy-constructor is private.

I thought the patches for using AbstractRRset::toText() in the branch is
some sort of a reasonable workaround for now, to make it work on that
platform. We cannot really ask users to patch their system Boost.  But
there're more issues in lexical_cast:

2. It is also broken in doing lexical_casts of integer and floating
point values. Element.from_and_to_json unittest fails, and if
lexical_cast is this broken, we cannot use it.

http://git.bind10.isc.org/~tester/builder/BIND10/20130913122501-Solaris10-sparc-Sunstudio/logs/unittests.out

We need to look at lexical_cast's implementation and if it does
something specific to SunStudio, or if it's wrongly using a packaged
version that has customizations, or something else is broken.  The
branch is not in review yet.

		Mukund
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 2881 bytes
Desc: not available
URL: <https://lists.isc.org/pipermail/bind10-dev/attachments/20130916/faea7539/attachment.bin>