Re: gcc-patches Digest 2 Apr 2000 04:45:49 -0000 Issue 740

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Tim" == Tim Josling <tej@melbpc.org.au> writes:

    Tim> This is the 'stick' approach.

It's certainly no fault of the Java team that they started working on
GCC before we had GC; to blaim them for using the old system would be
nothing short of ludicrous.

    Tim> I would suggest that as a 'carrot' complementary to this,
    Tim> documenting the garbage collection in the texinfo, and
    Tim> bringing the GC code uniformly up to GNU coding standards (eg
    Tim> ggc-simple.c - I have sent the author a detailed list of
    Tim> exceptions) would be an excellent idea.

The GC stuff is actually considerably simpler than, say, the largely
undocumented interface by which a front-end passes tree structure to
the back-end for conversion to RTL.  And the code quality of
ggc-simple.c, despite whatever nits you've come up with, is
considerably higher than many other parts of the compiler.  I'm not
saying more comments wouldn't help; only that, in my opinion, the code
there is considerably easier to read and make sense of than some other
parts of the system.

However, your point is well-taken.  Contributions for additional
documentation for the GC, as for any of the other internals in GCC,
would certainly be welcome!

In the short run, I'm sure that if the Java folks need a pointer or
two on how something works, someone will be happy to answer their
questions.  Hopefully, that answer will come in the form of additional
comments and/or documentation.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com