This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: demengler in libstdc++ is changed


Peter Barada <peter@the-baradas.com> writes:

| >| > Many C++ runtime implementors have reached the conclusion that the C
| >| > runtime is far more efficiently and effectively implemented on top of
| >| > C++ (rather than the inverse); however that idea seems to meet
| >| > (non-technical) resistance (I have no workable theory of why). 
| >| 
| >| Well, one obvious reason is that not all C programmers want to learn
| >| C++.
| >
| >I fully understand that.  But, I'm not requesting C programmers to
| >learn C++.
| 
| If the C runtime is in C++, then you are requiring that a C++
| compiler preexists to build the C compiler with.  That creates a whole
| new world of hurt when anyone tries to bootstrap GCC onto a new chip. 

  (1) if you don't like building a cross-compiler, consider that you
      can start with a C compiler and build a functioning GCC that way
      contain parts written in C++. 
  (2) anyone who is going to implement the C++ runtime would have to
      supply for (one way of the other) the C runtime.  And if you
      think hard about it have worked hard on it, you'll realize that
      what I said above is triviality even if it takes time to realize.

      The real question is how and when to make it happen.

| As it is now, even building the C++ cross compiler for existing
| environments is a PITA.  Why make it worse?

-- Gaby


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]