This is the mail archive of the libc-alpha@sources.redhat.com mailing list for the glibc project.


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

Re: GCC vs GLIBC: why this stance, Drepper ?!?


On Sun, Jul 01, 2001 at 10:44:06PM +0200, Martin v. Loewis wrote:
> 
> 1. You want libgcc_s be installed in /lib. This is possible through
>    --with-slibdir=/lib.

Not true.

> 
> 2. You are concerned that the API of libgcc_s will change in
>    backwards-incompatible ways. Different GCC developers have promised
>    that this will not happen.

I accept their promise.

> 
> 3. You want a documented libgcc_s API. I doubt any GCC contributor
>    would disagree that this is desirable; it's just that no volunteer
>    has stepped forward to provide this documentation.

It is a very nice thing to have.

> 
> 4. You are concerned that multiple installations of libgcc_s on a
>    single system must be avoided at all consts. While many have heard
>    you, it seems that nobody really agrees; for all the examples
>    you've given, an easy diagnosis of the problem will be given by the
>    dynamic linker, and an easy solution is available (setting
>    LD_LIBRARY_PATH in some cases, removing the extra libgcc_s in
>    others).

By default, only the system shared libgcc library should be used at
the run time if it is available. The user have to explicitly override
it. At this point, it is my only major concern on gcc 3.0.

> 
> 5. You want that libgcc_s is not used on Linux; instead, you want to
>    provide its interface through glibc. At the moment, it appears that
>    all this would achieve is to delay gcc releases much longer, and to
>    lockstep glibc and gcc releases in unacceptable ways.

No.

> 
> 6. You want that GCC authors to recognize that libgcc_s is a system
>    library. Not sure what this would improve technically, but I'll
>    happily recognize that.

See #4.

> 
> 7. You want that glibc maintainers are consulted in all libgcc_s
>    matters. Given that you and Ulrich have voiced their opinion
>    already, I don't see what further consultation would accomplish.

See #4.


H.J.


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