Problems with __cxa_demangle and the verbose terminate handler

[Mark Mitchell]

Matt Austern wrote:

> On Feb 22, 2004, at 11:01 AM, Mark Mitchell wrote:
> 
>> Gabriel Dos Reis wrote:
>>
>>> When called from "C" land, those containers should probably be
>>> instantiated with an allocator that either calls malloc() or use the
>>> nothrow form of new.  Useing "new" in  itself is not the problem, it
>>> not is calling the nothrow-form of new that is problematic.
>>>
>> I considered and rejected the use of a special allocator.
>>
>> If std::vector tries to allocate dynamic memory using an allocator 
>> that does not use throw an exception (but instead returns NULL), what 
>> is std::vector going to do?
>>
>> I believe std::vector it will then either crash or throw an exception, 
>> depending on the implementation.
>>
>> (In fact, I do not believe that [lib.allocator.requirements] allows an 
>> allocator's "allocate" function to return NULL.)
>>
>> If the new demangler called "operator new" directly, then it could 
>> indeed notice the return value being NULL and take appropriate action.
> 
> 
> What would the appropriate action be?

Return NULL, and set the status parameter to indicate memory allocation 
failure, as required by the ABI.

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com