C++ compiler situation (was Re: [grave] libstdc++2.8 needs versioned dependencies)
On Tue, May 12, 1998 at 01:48:40PM -0300, Adam P. Harris wrote:
> In message <[🔎] 19980512185214.58457@wi.leidenuniv.nl> you wrote:
> >On Tue, May 12, 1998 at 12:39:48PM -0400, Adam P. Harris wrote:
> >> Herbert Xu <herbert@gondor.apana.org.au> writes:
> >> > I think we should backout egcs and libstdc++2.8 from hamm and go back
> >> > to the old g++. For one, altg++ doesn't even work with it.
> >>
> >> Just to throw more fuel on this rather old fire, when I compile jade
> >> with egcs's g++, the program tends to dump core rather regularly. By
> >> backing out to g++272 vintage, the problem went away.
> >
> >We now have:
> >- GNU gcc 2.7.2.3
> >- egcs gcc (1.0.3-prerelease)
> >- egcs gobjc (1.0.3-prerelease)
> >- egcs g++ (1.0.3-prerelease)
> >- GNU g++272 [*]
> >- libg++272 (for GNU g++)
> >- libstdc++28 (for egcs g++) (1.0.3-prerelease)
> >- libg++28 (for egcs g++) (works with 1.0.3-prerelease)
> >
> >And RSN we'll have
> >- egcs-docs (containg the info files for egcs gcc)
> >- the iostream docs in libstdc++2.8-dev
> >
> >[*] Currently in slink, as nobody has bothered to provide feedback about it.
> > If someone provides me with positive feedback that having this package
> > in frozen would really be needed to compile a package for frozen,
> > I'd be more than willing to request it be moved there.
>
> Well, if you look at /usr/doc/jade/buildinfo.Debian, you shall see that
> I am well aware of this package and it's what I used to build jade
> after backing off from egcs:
OK. I'll submit a bug to get g++272 moved to frozen.
> >What more do you want?
>
> Ray, I am simply transmitting my experiences! Please don't jump down
> my throat.
I'm sorry, but the combination of a heat wave and the phrase "throwing fuel
on this old fire" did cause me not to be very happy with your message; I
took it as support for the idea that we should switch back to gcc 2.7.2.3
for all C++ packages, which would IMO be a grave mistake that would cause
an unsuccesful end to the current code freeze.
> What I would like to know, actually, is whether I should be notifying the
> jade upstream maintainer (James Clark) that his C++ is not up to snuff,
> and here's why, or alternatively, submit a bug against egcs. I think
> someone would have to look through the jade source to really determine
> that however.
If you've build jade using "-g", the core dump should provide clues on what
is the matter. I'm not a C++ programmer, so I don't think I can offer more
useful suggestions. You might want to ask on the egcs list directly.
On Tue, May 12, 1998 at 02:30:44PM -0400, Christopher C. Chimelis wrote:
> > By the way, g++272 is very excellent and nice and it was bascially
> > smooth as silk to install it and use it. Works great!
>
> Hmmm...well, I feel Ray's pain here since I'm one of the people pressuring
> him to oust g++272 due to it being outdated and not multiple architecture
> friendly.
Currently, my standpoint is that egcs g++ should be used on all
architectures to compile all C++ packages. For cases where it is not
possible for a maintainer to get a C++ program to compile & work with egcs
g++, I created the g++272 package.
The g++272 package is derived from our gcc-2.7.2.3 source; if there's
anything I can do to get it to build easier on non-intel architectures,
please let me know.
> On the Alphas at least, g++272 will not build very easily and, even if we
> get it to, it doesn't support alot of the features of libstdc++ from egcs.
Yes. It's intended only for those cases where egcs g++ cannot be used yet.
> I suppose we could split "the tree", so to speak, when it comes to other
> architectures, but I think Ray is just trying for some unity (and
> therefore personal sanity on his part) amongst the architecture ports for
> something so integral as a fundamental library and compiler.
I'm not the maintainer for egcs, gcc, gobjc, libstdc++, libg++ etc. Because
these packages are so fundamental to building a proper release, I've been
doing non-maintainer uploads (with a lot of help from Matthias Klose and
folks on #debian) of most of them in order to provide as much flexibility in
them without making fundamental changes (which I think should not be made
during the code freeze) (making libstdc++28-dev and libg++272-dev coexist
would be too big a change, so unfortunately those who need both egcs g++ and
g++272 with their libraries, need to switch between the library -dev
packages).
Greetings,
Ray
--
PATRIOTISM A great British writer once said that if he had to choose
between betraying his country and betraying a friend he hoped he would
have the decency to betray his country.
- The Hipcrime Vocab by Chad C. Mulligan
--
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: