[1.37][Function] Request permission to merge changes
Aleksey pointed out that Boost.Function in 1.36.0 has a critical thread-safety problem. I've now fixed the problem on the trunk, and---after regression tests cycle---would like to merge these changes (along with other long-stable fixes to Function) to the release branch. Okay? - Doug
On Tue, Oct 14, 2008 at 11:40 AM, Doug Gregor <doug.gregor@gmail.com> wrote:
Aleksey pointed out that Boost.Function in 1.36.0 has a critical thread-safety problem. I've now fixed the problem on the trunk, and---after regression tests cycle---would like to merge these changes (along with other long-stable fixes to Function) to the release branch. Okay?
Yes. --Beman
Doug Gregor wrote:
Aleksey pointed out that Boost.Function in 1.36.0 has a critical thread-safety problem. I've now fixed the problem on the trunk, and---after regression tests cycle---would like to merge these changes (along with other long-stable fixes to Function) to the release branch. Okay?
Yes, we really need this fix. Can you describe the nature of the other changes you plan to merge? -- Eric Niebler BoostPro Computing http://www.boostpro.com
On Tue, Oct 14, 2008 at 11:57 AM, Eric Niebler <eric@boost-consulting.com> wrote:
Doug Gregor wrote:
Aleksey pointed out that Boost.Function in 1.36.0 has a critical thread-safety problem. I've now fixed the problem on the trunk, and---after regression tests cycle---would like to merge these changes (along with other long-stable fixes to Function) to the release branch. Okay?
Yes, we really need this fix. Can you describe the nature of the other changes you plan to merge?
Of course. They are: - const-correctness for function::target(): https://svn.boost.org/trac/boost/ticket/736 - performance improvement for function::swap(): https://svn.boost.org/trac/boost/ticket/1910 - added function_fwd.hpp header: https://svn.boost.org/trac/boost/ticket/1668 - added typeof support: https://svn.boost.org/trac/boost/ticket/1668 - workarounds for the latest CodeGear C++ Builder: https://svn.boost.org/trac/boost/ticket/2353 It is *possible* to make the thread-safety change on the release branch without bringing over the other changes, but everything else has been on trunk for nearly a month without any breakage. So, I think it's probably riskier to try to bring fixes over piecemeal than to just do a merge from trunk's Function. - Doug
Doug Gregor wrote:
On Tue, Oct 14, 2008 at 11:57 AM, Eric Niebler <eric@boost-consulting.com> wrote:
Doug Gregor wrote:
Aleksey pointed out that Boost.Function in 1.36.0 has a critical thread-safety problem. I've now fixed the problem on the trunk, and---after regression tests cycle---would like to merge these changes (along with other long-stable fixes to Function) to the release branch. Okay? Yes, we really need this fix. Can you describe the nature of the other changes you plan to merge?
Of course. They are:
- const-correctness for function::target(): https://svn.boost.org/trac/boost/ticket/736 - performance improvement for function::swap(): https://svn.boost.org/trac/boost/ticket/1910 - added function_fwd.hpp header: https://svn.boost.org/trac/boost/ticket/1668 - added typeof support: https://svn.boost.org/trac/boost/ticket/1668
Wrong link here. What changes did you have to make for this one?
- workarounds for the latest CodeGear C++ Builder: https://svn.boost.org/trac/boost/ticket/2353
It is *possible* to make the thread-safety change on the release branch without bringing over the other changes, but everything else has been on trunk for nearly a month without any breakage. So, I think it's probably riskier to try to bring fixes over piecemeal than to just do a merge from trunk's Function.
I tend to agree. -- Eric Niebler BoostPro Computing http://www.boostpro.com
On Tue, Oct 14, 2008 at 1:42 PM, Eric Niebler <eric@boost-consulting.com> wrote:
Doug Gregor wrote:
- added typeof support: https://svn.boost.org/trac/boost/ticket/1668
Wrong link here. What changes did you have to make for this one?
Argh! It's just a bunch of BOOST_TYPEOF_REGISTER_TEMPLATE uses, in the new boost/function/function_typeof.hpp. https://svn.boost.org/trac/boost/ticket/1621 - Doug
Doug Gregor wrote:
On Tue, Oct 14, 2008 at 1:42 PM, Eric Niebler <eric@boost-consulting.com> wrote:
Doug Gregor wrote:
- added typeof support: https://svn.boost.org/trac/boost/ticket/1668 Wrong link here. What changes did you have to make for this one?
Argh! It's just a bunch of BOOST_TYPEOF_REGISTER_TEMPLATE uses, in the new boost/function/function_typeof.hpp.
Good, just making sure function.hpp isn't pulling in the typeof stuff. OK to merge. -- Eric Niebler BoostPro Computing http://www.boostpro.com
On Thu, Oct 16, 2008 at 10:19 AM, Doug Gregor <doug.gregor@gmail.com> wrote:
On Tue, Oct 14, 2008 at 3:02 PM, Eric Niebler <eric@boost-consulting.com> wrote:
Good, just making sure function.hpp isn't pulling in the typeof stuff. OK to merge.
Everything is fine on the trunk, so I've merged the Boost.Function bits.
Thanks, --Beman
On Tue, Oct 14, 2008 at 9:40 AM, Doug Gregor <doug.gregor@gmail.com> wrote:
Aleksey pointed out that Boost.Function in 1.36.0 has a critical thread-safety problem. I've now fixed the problem on the trunk, and---after regression tests cycle---would like to merge these changes (along with other long-stable fixes to Function) to the release branch.
Can we expect a hotfix for 1.36? <crosses fingers> Jon
participants (4)
-
Beman Dawes -
Doug Gregor -
Eric Niebler -
Jonathan Franklin