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: [Ada] Patch to fix bug reporting instructions (3.1 branch)


> For C, libiberty gets compiled with the bootstrap compiler, then does not
> get recompiled by GCC for the subsequent stages of bootstrap.  There's
> been the occasional issue with this on systems supporting multiple ABIs /
> binary formats, but in general it works - which seems fairly similar to
> the situation with Ada.

It is not just a matter of multiple ABI's. There are also cases where gcc
is not fully 100% compatible with the ABI for the vendor compilers. There
have been several threads in the past discussing such cases. 

To me these cases are a little bit similar, and it seems a mistake not to
recompile libiberty with a consistent compiler.

The cases however are not that similar. I assume that Joseph is familiar
with the detailed internal structure of GNAT ... the point is that GNAT
is a mixture of C and Ada that is very intimate, intimate enough for 
example that it is out of the question to try to use an Ada compiler other
than GNAT for compiling GNAT, since the C part knows a lot about implementation
dependent details of GNAT (not to mention the fact that GNAT itself uses a lot
of GNAT dependent features). This means that it is really quite important
to use compatible versions of GNAT and C, and it is surprising if incompatible
versions work.

I agree with Richard that it was a mistake to support the gnatgcc case,
since this is asking for trouble, but I also agree that it is too late
to remove it.

I am a bit at a loss to understand why this is a big issue. I don't know
of any Ada user who is concerned by this right now, and there are far more
important fish to fry on the GNAT front in terms of getting it fully
operational under GCC 3 (not least of which is correcting some of the
remaining errors in code generation on some targets).

Joseph, I assume you are an Ada user (otherwise I have trouble understanding
this intense interest in this one issue, can you be more clear as to why you
think this is such an important issue?)


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