Re: A completely different approach to EH runtime

["H . J . Lu" <hjl at valinux dot com>]

On Wed, Feb 21, 2001 at 08:05:37PM -0300, Alexandre Oliva wrote:
> On Feb 21, 2001, "H . J . Lu" <hjl@valinux.com> wrote:
> 
> > On Wed, Feb 21, 2001 at 07:40:27PM -0300, Alexandre Oliva wrote:
> 
> >> I'd advise people to install other versions of GCC in other
> >> directories.  I suppose they won't be building programs used during
> >> start-up with this new version of GCC.
> 
> > Do you think it will encourage Linux people to try new gcc?
> 
> It won't be a problem, as long as they don't replace start-up
> programs.  Why did you choose to skip that part of the argument?

Are you prepared to deal with 2 shared libgcc visible to ld.so at
the same time and one of which doesn't come from the system vendor?

> 
> IMO, it is radically different, unless you use GCC to build system
> programs, and replace their original versions with the ones you've
> built with this new GCC.  Joe Hacker won't do that.
> 
> Can't one install glibc in a separate directory and use it only for a
> few programs?  That's just the same as installing GCC in a separate
> directory, and arranging for libgcc_so to be used only with the
> programs built with that particular GCC.
> 

As a glibc developer, I can tell you that it is not safe unless you
know exactly what you are doing. You can't have 2 shared glibc
visible to ld.so at the same time. Think about compiling DSOs with
the new gcc and all the issuses related to dlopen. When you have
2 shared glibc or libgcc, it may be almost impossible to know which
one will be used at the run-time and which one should be used. That
is why we have one shared libc which is 100% backward compatible with
the existing dynamic binaries.


-- 
H.J. Lu (hjl@valinux.com)