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: patch for -Wno-long-long and early GNAT compilers

> <<Please don't split hairs.  We went over this at length last year; the
> Ada front end in the released GCC 3.1 has to be buildable with the
> commonly available gnat-3.13.  And I want to go on record as saying
> that it *should* be buildable with *any* standard-compliant Ada
> compiler.
> >>

Robert Dewar writes:
> Just to address the last sentence here in more detail.
> GNAT is entirely full of assumptions that it is being built with itself,
> it freely uses GNAT extensions in the areas of pragmas and attributes,
> and indeed some of these pragmas and attributes are designed specifically
> to be useful for the compiler itself. It also uses library units
> which themselves rely freely on GNAT'icisms.

OK, clearly it isn't going to happen that GNAT will become buildable
by any Ada compiler, and it doesn't seem that there is a real reason
that we should care.

But what about older GNAT versions?  For example, Debian "potato" has
version 3.12p; Debian "woody" has 3.13p.  I'm sure that one or both of
these versions have been packaged for many other OSes as well.

It would be easier for third parties to build from source if one or both
of those versions could be made to work.  Is this out of the question as
too difficult, or is it a more minor matter?  I'm not saying that this
should be a requirement, but I'd like some clue as to the cost.  Maybe
it's good enough that just 3.14 can do the job.

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