[Gc] Re: Memory leak located!

Christophe Meessen meessen at cppm.in2p3.fr
Tue May 27 04:52:23 PDT 2008


I missed that  info and was not aware of the previous discussions. I 
looked at the only documentation only and what I could read in some files.

Using gc_allocator is ok as long as we are in control. The other good 
thing is that we know what is going on. But the difficulty is that some 
allocations are out of our controle, like strings in exceptions for 
instance. There is thus a limit with this strategy.

Do you mean that with the suggested change all new() operations will 
create collectable blocks ? Can this be done with a compiler flag ?

Thus will push back some source of problems but will only work with 
block allocation taking in account the modified new call. In general the 
template classes will do fine. But there are other classes where this 
won't work.

It looks like a close integration with the compiler and language is 
really needed (ala managed code).

Bien cordialement,

Ch. Meessen

More information about the Gc mailing list