Bug in egcs-1.1b (egcs-2.91.57)

[Alexandre Oliva]

John Polstra <jdp@polstra.com> writes:

> The following test case illustrates a bug in egcs-1.1b (egcs-2.91.57).
> The compiler gets confused about the types of things and erroneously
> complains about the assignment "spec.handler = &HandleReadCommand;"

Fixed in the latest snapshot.  BTW, in order to form a pointer to
member, you *must* use a qualified name such as
&DeviceHandler::HandleReadCommand.  Just &HandleReadCommand won't do,
even within a member function of DeviceHandler, according to the C++
Standard.  Furthermore, your program will not compile with -ansi
-pedantic; it's missing the `typename' keyword before
template-dependent type names

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil