[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: How to change a symlink without breaking anything (problems with g++)



On Tue, Jul 01, 2003 at 08:14:34PM +0200, Morten Eriksen wrote:
> (In case someone is interested, I'm asking because I just stumbled
> over a surprising issue with the g++ packages: doing an
> 
>        # apt-get install g++
> 
> installed g++-3.3 (I was expecting just an upgrade to the previously
> installed g++-2.95), and made /usr/bin/g++ into a symlink pointing to
> it.

Why is this surprising?  This isn't happening in Woody, so you're either
using Sarge or Sid.  The C++ transition has endlessly been debated and
beaten to death in a variety of public fora... if you don't know about it,
you're living under a rock and probably shouldn't be using anything other
than stable.

> Just switching g++ versions like this is bound to cause mayhem for
> all development against any C++ libraries on the system already
> compiled with g++ v2.95, so I'm surprised that this happens just like
> that from the "testing" branch..?

Yep.  Amazingly enough, Debian planned this transition, and packages
needing to move to the new ABI are doing so.  Perhaps you haven't noticed
all the packages now with "c102" in their names?  Sid is pretty much
complete, I think... Sarge probably has a way to go.  I've not updated my
Sarge box in a while so I don't know for sure.

Locally compiled applications, of course, have the potential to break.
OTOT, if you're compiling stuff locally, the assumption is that you have at
least minimal clue.

> Removing the g++-3.3 package doesn't
> work either, as apt-get seems to now refuse to remove or install
> _either_ g++-2.95 or g++-3.3 -- it insists on having both? Weird.)

So?  The compilers are independent... use 2.95 when you need to and use 3.3
when you need to.  Building kernels, for example, should not be done with
gcc 3.3 just yet (yes, I'm sure there's some idiot who's about to rant
about how wonderful his gcc_3.3-compiled kernel is... pay attention to lkml
for the number of regressions still being found).

Most of the rest of the world started this process some time ago.

-- 
 Marc Wilson |     "Confound these ancestors.... They've stolen our
 msw@cox.net |     best ideas!"  - Ben Jonson



Reply to: