Re: gnat-3.10

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Corey says

<<What they do not seem to do, which disturbs me a little, is distribute
the GCC base they use to build their stuff on their FTP site.  Since
the patches they supply don't apply perfectly, they are obviously
using a patched GCC base and do not provide information about how to
get it.  But perhaps they are relying on b) above in the GPL for that
specific part, which would be perfectly fine legally.  I'd rather have
it on the FTP site, though.
>>

Generally the patches we supply *should* apply cleanly to the base. That
is certainly the intention. When it is not the case, it is a glitch. Right
now, the patch situation is quite messy, because of the very large delta
between 2.7.1 and 2.8. Virtually all of our gcc patches are incorporated
into 2.8, but not all the 2.8 patches are incorporated into our patch
file, only the ones that Ada needs. Keeping this straight is not easy,
and that is especially true in the case of the many different targets
we support.

We are expecting gcc 2.8 to be released very soon, and this should
greatly simplify the issue of getting the right version of gcc sources
and patches (at least for a transitory moment, the GNAT patch file
will become empty or very close to it).

The current gcc-272.dif is nearly 6000 lines long (this file originall
started out as a very small file of a few critical patches). Furthermore,
things are more complicated now because of the appearence of the 2.7.2.1
release (this was a kind of semi-official release demanded by the Linux
community to fix one not very important bug that got blown out of
proportion), so now we need a gcc-2721.dif as well.

As I say, these problems should completely disappear when gcc 2.8 is
issued. One of the good things here is that all GNAT patches do get
reflected in the main GCC sources that FSF develops almost immediately
(that's because the developer of these patches, and the maintainer
of the FSF version of GCC are the same person :-)