Shared library annoyance with gcc-3_0-branch

Daniel Jacobowitz dmj+@andrew.cmu.edu
Sun Feb 18 21:51:00 GMT 2001


On Sun, Feb 18, 2001 at 09:37:01PM -0800, Mark Mitchell wrote:
> >>>>> "Daniel" == Daniel Jacobowitz <dmj+@andrew.cmu.edu> writes:
> 
>     Daniel> For Debian, at least, I'm going to have to recommend that
>     Daniel> we not build a shared libgcc.  IIRC, there is an option to
>     Daniel> build other libraries shared, and maintain a static libgcc
>     Daniel> - right?
> 
> Why not build the library (so that users that want it can use it)
> but change the GCC driver to not use it by default, if you think that
> will be best for your users?
> 
> If you do not build it, you're making it impossible for people to
> correct semantics in some cases.

I agree, this sounds like a better idea.

>     Daniel> This may be an already-hashed-out question, whose answer I
>     Daniel> just lost in the confusion... for programs which are
>     Daniel> neither shared libraries themselves nor written in
>     Daniel> C++/Java/etc, is there any correctness loss in a static
>     Daniel> libgcc?  Is this feasible?
> 
> There should be no correctness issue for C programs.
> 
> Note that for C programs (as opposed to C shared libraries), we still
> use -static-libgcc by default.  So, it maybe that we already do what
> you want.

Will this behave correctly if glibc explicitly does not link to
libgcc_s?  That seems like it would defeat the purpose of of having a
shared libgcc fairly thoroughly.  My goal is to get C programs, at
least those which require no shared libraries beyong the standard C
library, to run without shared libgcc being available; I consider this
a robustness issue.  It may not be feasible, though.

Dan

/--------------------------------\  /--------------------------------\
|       Daniel Jacobowitz        |__|        SCS Class of 2002       |
|   Debian GNU/Linux Developer    __    Carnegie Mellon University   |
|         dan@debian.org         |  |       dmj+@andrew.cmu.edu      |
\--------------------------------/  \--------------------------------/



More information about the Gcc mailing list