On Mon, Oct 04, 2010 at 11:08:53AM +0200, Carmelo AMOROSO wrote:
>Folks,
>time ago we added argp support into ST branch of uClibc because it is
>used by some elfutils tests and prelink tool that we build (we did not
>want to modify them to use getopt instead of argp). So we ported the
>argp support from glibc into uClibc and modified it accordingly to fit
>into the uClibc build system.  We did not upstream this work because it
>is a feature that is almost never used, so it does not worth to add this
>extra code into uClibc (I discussed this with Bernhard)
>
>If some of you are going to use prelink, this support will be actually
>required and should have our patches. I was thinking to create a
>"libuargp" to keep the argp code in a separate shared object avoiding to
>populate the libc.so with seldom used functions.
>
>So "libuargp" would be transparently linkedby using the AS_NEEDED
>directive in the linker script libc.so for those applications that
>actually requires argp symbols (as for libubacktrace).
>
>What do you think ?

Can you provide a patch for us to look at?

As said previously a libuargp is fine with me for it doesn't penalise
the common case except taking up storage-space in a few cases (iff
prelink-support is enabled). If there is consensus that we can build it
in if prelink is enabled and have it AS_NEEDED otherwise, then it's fine
with me.
_______________________________________________
uClibc mailing list
[email protected]
http://lists.busybox.net/mailman/listinfo/uclibc

Reply via email to