Antony Polukhin wrote:

2011/5/19 Jeroen Habraken <vexocide@gmail.com>:

...

Hi,

Whilst updating SVN today I spotted a change to boost::lexical_cast, https://svn.boost.org/trac/boost/changeset/71958, which introduces a try { } catch(...) { } construction. I'm curious as to why this catch-all is needed since it seems to be there to deal with boost::numeric_cast exceptions and whether it could lead to a std::bad_alloc being rethrown as a bad_lexical_cast.

numeric_cast does not use new operator, so std::bad_alloc will not be rethrown as a bad_lexical_cast. But you are right, and it will be more appropriate to catch only boost::numeric_cast exceptions. I`ll fix it in next commit. Great thanks for reporting this.

Hi, I see new test have been added. Have you run the new tests with the previous lexical cast version? Best, Vicente -- View this message in context: http://boost.2283326.n4.nabble.com/Generic-try-catch-in-boost-lexical-cast-t... Sent from the Boost - Dev mailing list archive at Nabble.com.

Antony Polukhin wrote:

2011/5/20 Vicente Botet <vicente.botet@wanadoo.fr>:

...

Hi,

I see new test have been added. Have you run the new tests with the previous lexical cast version?

Yes, I have. They passed successfully.

Vicente, you have commented some time ago a ticket https://svn.boost.org/trac/boost/ticket/4946 If you have some time and still can reproduce this bug, please check the fix. It is just about adding #include <boost/utility/value_init.hpp> and using boost::value_initialized<Target> result; .

Hi, I have never tried to reproduce the tis bug. It was just that the proposed patch introduced a new warning. Now that you have changed the patch there is no reason this specific warning appears. Best, Vicente -- View this message in context: http://boost.2283326.n4.nabble.com/Generic-try-catch-in-boost-lexical-cast-t... Sent from the Boost - Dev mailing list archive at Nabble.com.