Re: --disable-versioning and --enable-oldest-abi

[Andreas Jaeger <aj at suse dot com>]

On 07/07/2013 07:19 PM, Joseph S. Myers wrote:
> The --disable-versioning configure option has long been bitrotten (I don't 
> know of any glibc bug report, open or closed, for this issue).  (If you 
> use it, you get a configure warning saying
> 
> *** WARNING: You should not compile GNU libc without versioning. Not using
> *** versioning will introduce incompatibilities so that old binaries
> *** will not run anymore.
> *** For versioning you need recent binutils (binutils-2.8.1.0.23 or newer).
> 
> and I suspect it's been bitrotten since not long after binutils versions 
> of that vintage were current.)
> 
> There are some changes in EGLIBC to make this option work.  Those are also 
> by now bitrotten (that is, --disable-versioning doesn't work there 
> either).  <http://www.eglibc.org/archives/patches/msg00631.html> is the 
> first version of those changes, with analysis of the various forms of 
> breakage (there are of course differences from the variant now in EGLIBC).
> 
> I don't think it makes sense to keep this configure option in broken form; 
> we should either fix it (possibly along the lines of the changes in 
> EGLIBC) or remove it.  Given the limited utility of the option and the 
> likelihood of such a deviation from default configuration breaking in 
> future, my inclination (for after 2.18 branches, of course) is removal.  
> Comments?

No bug reports and a broken support shows indeed that nobody is using
it. I'm not aware of any current reasons for using it but if somebody
has some, please speak up so that we understand the usecase.

I'm fine with removing it.


> 
> The --enable-oldest-abi configure option, another obscure ABI-breaking 
> option (not even documented, unlike --disable-versioning), is also broken 
> (bug 6652) (a test for 32-bit x86 with --enable-oldest-abi=2.3 shows the 
> breakage; the cases I tried for x86_64 didn't fail the build although I've 
> no idea if the generated libraries worked).  I'm inclined to think this 
> option too should be removed, although I think that wouldn't actually 
> bring much cleanup beyond removing the configure option (whereas there are 
> plenty of conditionals that could be removed if --disable-versioning is 
> removed).  Comments?

I wonder why would somebody want to really use this,

Andreas
-- 
 Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
  SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126