Re: [PATCH] Provide aliases for backward compatibility

[Christopher Roy Bratusek <zanghar freenet de>]

Am Tue, 8 Jun 2010 10:49:20 +0400
schrieb "Alexey I. Froloff" <raorn altlinux org>:

> On Tue, Jun 08, 2010 at 10:46:55AM +0900, Teika Kazura wrote:
> > Oh, oh, Alexey, thanks for your hard work, but these functions are not
> > available from C. They're not declared in any librep header
> > files. (Read /usr/include/rep/*.) They're extern, but can be called
> > only from inside of librep.
> These symbols are exported and can be used by binaries linked
> against librep.so.9.  If you delete these symbols without
> relinking all dependent binaries, you'll get load error.
> 
> If these symbols can't be used outside of librep, no need to
> export them.  Otherwise, provide necessary aliases or change
> SONAME.
> 
> > Furthermore, these functions are introduced in this January, and the
> > only published version with them is 0.90.5, so I guess it's virtually
> > harmless without your patch.
> You know, in ALT Linux librep is used by one package and a half,
> but I thought binary compatibility is a good thing in general.
> 
> P.S. http://people.redhat.com/drepper/dsohowto.pdf
				^^^^^^^ I'm not going to read that.

Well, -p on those was a mistake, so they have been removed, additionally it's been
the first commit right after 0.90.5, so most GIT-users won't even ever had -p
variants. Also it's only rep-gtk/sawfish linking against (pager is linking against
sawfish), so I haven't thought that this would be necessary.

Either way SONAME bump is prefered over aliases, they're only polluting the code in
my opinion.

Chris


-- 

GlibC Bugreport + GlibC Fix = NOTABUG, do not re-open (UD)

Re: Embedded-Plattformen und RedHat/C...
Gesendet von Peter am Mi, 6. Mai um 20:53

Falls dieser Mann - hoffentlich in ferner Zukunft - einst in der Kiste liegt,
wird man auf seinen Grabstein schreiben: "Do not reopen."