This is the mail archive of the 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: [Ada] Patch to fix bug reporting instructions (3.1 branch)

On Wed, 1 May 2002, Geert Bosch wrote:

> Trying to autodetect various "gnatgcc" installations on GNU releases
> is asking for trouble. We should just have one reliable compiler 
> distribution to bootstrap from, and that is it. Later, we can refer
> to using GCC 3.(X-1) for bootstrapping GCC 3.X.

Do you mean people using their own installation of 3.1 to bootstrap 3.2,
or do you mean an ACT-provided binary every time?  If it's OK for people
to use their own 3.1 installation, why not their own 3.1[34] installation?
What about their own installation where 3.1 is the Ada compiler but the C
compiler is a more conservative version, such as 3.0?  The GNU/Linux
distributions often choose different compiler versions for different
languages.  Again, RMS has stated that it's important for GNU Ada to
support GNU/Linux well.

Could the SC please consider this issue?  The questions for consideration

* How far should GCC go in detecting a broken or too old GNAT installation
at configure time, and disabling the Ada front end if the GNAT
installation is unsuitable for building it?

* Should bootstrapping the Ada front end in GCC 3.1.x be supported using
previous installations of GNU Ada that do not use ACT-provided binaries?
For example, those included in GNU/Linux distributions, or where someone
has started from the ACT binaries for some version of GNAT and used them
to rebuild from sources that same version of GNAT (possibly then building
a series of subsequent versions of GNAT, each from sources, using the
installed previous version)?  If so, should this be supported where the C
and Ada compilers are different (since 2.8.1 is reported not to work well
for C with current glibc)?

The disagreement seems to be largely between the ACT developers used to
one fairly uniform method of development (current ACT GNAT installed as C
and Ada compiler, used to build next version) and the other (mainly
non-Ada) developers from the wider free software world, where portability
to a wider range of user environments is considered more important and
older compiler versions are commonly used for longer for stability, who to
a greater extent consider that needing to use outside binaries rather than
sources plus existing binaries for a build is to be minimised, and would
prefer GNU Ada to operate more like other front ends now it is an
integrated part of GCC.

One question that may need to be considered in future - and also may be of
importance to GNU/Linux distributors, though it would be of less
importance if every version were uniformly better and more stable than the
last, so GNU/Linux distributors had no occasion to stick to old versions -
is whether GCC 3.1 should be supported to bootstrap Ada in GCC 3.3?

Joseph S. Myers

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