Missing ns_initparse

Peter Mazinger ps.m at gmx.net
Sat Jul 9 11:24:09 UTC 2011


> On Thu, Jul 7, 2011 at 12:04 PM, Daniel Mack <zonque at gmail.com> wrote:
> > Ok. Would it be an option then to add it? Reading the guidelines of
> > this project, I'm not sure whether support for this functions are
> > wanted at all because at least the implemenations found in NetBDS's
> > libc is not particularily small. What's the verdict here? Can things
> > be added as long as they're build conditionally? Are there any
> > alternative aproaches to achive support for these functions? I'm sure
> > connman is not the only program that fails to link against uClibc due
> > to this.
> 
> No opinions? So I dare to propose an implementation :)
> 
> Attached is patch that is not yet meant for inclusion but for RFC
> only. I can split it up into multiple smaller pieces once I know that
> it's basically ok.
> 
> Some parts of the patch move symbols from libc to libresolv, according
> to the header files they're defined in and also to bring it in sync
> with what glibc does.

I am not sure about moving symbols to libresolv, autoconf and friends should find the symbols in libc as well.

> Also, I haven't done a lot of testing yet, and major parts are just
> copied over from NetBSD's libc implementation and amended according to
> uClibc's internals.

lib*_hidden_* has to be used only if a function is reused within the same lib*.so, simply renaming libc_hidden_* is not OK (and adding it to all others is not needed).
You build first without any lib*_hidden* and check
with readelf -r lib*.so, lib*_hidden needs applied to all functions that
show up as jump relocation.

You have left some libc_hidden_* in, it does not make any sense in
files that go to another lib.

ns_name.c is "empty"

Peter

> 
> Please let me know what you think.
> 
> 
> Thanks,
> Daniel

-- 
NEU: FreePhone - kostenlos mobil telefonieren!			
Jetzt informieren: http://www.gmx.net/de/go/freephone


More information about the uClibc mailing list