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]

Re: Ada (Was: GCC 3.0 Release Criteria)


Joe Buck <jbuck@possibly.synopsys.com> wrote:
> [...] Given GNAT's large size and the smaller number of interested
> people (compared to C and C++), and the fact that the bootstrap
> process cannot build it (the front end is written in Ada so you need
> an Ada compiler such as an earlier version of GNAT to build it) I
> don't think it's really that desirable to physically put the GNAT
> source into the GCC tree.

The upcoming Debian release has GNAT in it, and I believe members of
the GNAT/Linux team are talking with various other Linux distribution
guyes, this probably means that most GNU/Linux systems will have it
installed by default soon.

For most other (proprietary) systems, you must pay to have a closed
source C compiler to build GCC, so having to download a free GNAT
binary in the case you want to build GCC with LANGUAGES="c c++ ada" is
not a real problem IMHO.

> Rather, it would suffice to include pointers on the gcc web site and
> documentation to the corresponding information for GNAT, so that
> interested people could load it.

That what the situation is right now. 

What is the point of renaming GCC to the GNU Compiler Collection if
significant (at least in size and functionality) front ends are left
out of the public development process?

One benefit of having the Ada front end in is that its bootstrap and
runtime build are valuable tests of the back-end. Also, there is a
publically available huge validation test suite for Ada (formerly
known as ACVC), and also a performance test suite (ACES). Huge here
means thousands of tests.

Provided that Ada Core Technologies finds it possible to integrate the
GNAT sources in the GCC development process (not obvious), I volunteer
to work on packaging the available Ada test suites so they can be used
easily to test GCC (and depending on timing, for 3.0).

Background: I worked for ACT in the past and wrote most of the
automated build and test system used for GNAT that I believe is still
in use (and extended ;-). The ACVC testing is done by ACT daily, but
with the additional complexity of non executable tests and peculiar
build process and reporting to abide by the standard, which is not
needed per see.

-- 
Laurent Guerby <guerby@acm.org>

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