This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: per-port symbol exports (was Re: libstdc++ related boostrap failure)


On Sat, Feb 15, 2003 at 11:23:07AM -0600, Benjamin Kosnik wrote:
> 
> Also, could you sketch to me how you think they could be implemented,
> given the pending ld patch? I'm now curious (I remember this being
> constraining to me a while back.)

The part of my message that started with "second attempt" is a pretty good
example:  we have some special text in a comment, below which we optionally
insert some other text.

Instead of making a symlink to the basic linker-map.<style> as we do now,
we would unconditionally copy the basic map file into the build directory.
Then, optionally, do a bit of sed/ed to drop in a per-port file fragment.
(Could probably do it all at once, but seperate steps are easier to debug.)

As for choosing the fragment, probably something like

    additional_per_port_export_symbols=none
    ...

    [ configure.target maybe sets a_p_p_e_s=$srcdir/a/particular/file ]
    ...

    cp $srcdir/config/linker-map.<style> ./src/libstdc++-symbol.ver
    if test $a_p_p_e_s != none; then
        [ drop $a_p_p_e_s into ./src/libstdc++-symbol.ver ]
    fi


Phil

-- 
I would therefore like to posit that computing's central challenge, viz. "How
not to make a mess of it," has /not/ been met.
                                                 - Edsger Dijkstra, 1930-2002


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