[The message I am replying to was sent from a newsgroup and I am sending my
reply (this) as a message to the list.]
"Edward Diener"
What Mr. Dimov is saying is that it is impossible to place the code you would like to use above in Boost because it would affect other Boost libraries which do not need RTTI to work.
That's no problem. The Boost headers already have more than 400 (including comment) lines that test for the "BOOST_MSVC" preprocessor symbol. In other words, there is already an abundance of code that is relevant only to MS VC.
Furthermore if it were placed in the header file of a library which did need RTTI to work, just including that header file, without actually using code in it that needed RTTI such as dynamic_cast, would cause an immediate compile time error.
Perhaps. I hope, though, that the possibility that it might not be useful won't prevent someone from communicating the idea to someone familiar with the relevant headers, since there is a possibility that it might be useful. I have more confidence in the developers that a solution is possible. I might have missed something, but I don't understand why it is relevant to say that it won't work. It might not work, but it might work. Yet you are saying it definitely won't work. That seems innapropriate to me.
Samuel wrote:
[The message I am replying to was sent from a newsgroup and I am sending my reply (this) as a message to the list.]
"Edward Diener"
wrote in message news:cfmhi0$4fd$1@sea.gmane.org... What Mr. Dimov is saying is that it is impossible to place the code you would like to use above in Boost because it would affect other Boost libraries which do not need RTTI to work.
That's no problem. The Boost headers already have more than 400 (including comment) lines that test for the "BOOST_MSVC" preprocessor symbol. In other words, there is already an abundance of code that is relevant only to MS VC.
This is irrelevant. Many Boost libraries work fine under MSVC with RTTI "disabled". Note that under MSVC /GR- does not disable RTTI altogether; only polymorphic types are affected. For example, typeid(T) still works.
Furthermore if it were placed in the header file of a library which did need RTTI to work, just including that header file, without actually using code in it that needed RTTI such as dynamic_cast, would cause an immediate compile time error.
Perhaps. I hope, though, that the possibility that it might not be useful won't prevent someone from communicating the idea to someone familiar with the relevant headers, since there is a possibility that it might be useful. I have more confidence in the developers that a solution is possible.
I might have missed something, but I don't understand why it is relevant to say that it won't work. It might not work, but it might work. Yet you are saying it definitely won't work. That seems innapropriate to me.
The "easy" solution you proposed doesn't work because it's too coarse-grained.
----- Original Message -----
From: "Peter Dimov"
This is irrelevant.
I think it is more productive to dwell on what is relevant. Yes, I know that is vague, but I hope it is understood by enough people.
participants (2)
-
Peter Dimov
-
Samuel