Static linking considered harmful

[Tony Nelson]

At 11:34 AM +0100 11/22/06, Arjan van de Ven wrote:
>On Wed, 2006-11-22 at 11:24 +0100, Patrice Dumas wrote:
>> On Wed, Nov 22, 2006 at 05:08:00AM -0500, Jakub Jelinek wrote:
>> > Hi!
>> >
>> > We should more proactively discourage static linking in FC7+, for
>> > reasons for that see
>> > http://people.redhat.com/drepper/no_static_linking.html
>> >
>> > Removing libc.a would be most effective, but I'm afraid we still need
>> > a handful of statically linked binaries for boot time initialization and
>> > system recovery utilities.
>>
>> Not only. There are cases when all those issues are moot, a prominent one
>> being for numerical models. Compiling models statically makes it possible
>> to run them on any other linux (including different fedora version) box
>> without recompiling
>
>actually static linking DECREASES that portability !!

Users of numerical models often want bit-reproducable results.  It's not
necessarily stupid, as if the results are identical no further comparison
is needed, while if the results are merely "close", much thought must go
into deciding whether they are close enough.  Using different version math
libraries makes identical results less likely, and makes more work for the
model's users.  They would probably switch to a more useful distro, and
good riddance to them!  Thbpt!
-- 
____________________________________________________________________
TonyN.:'    The Great Writ     <mailto:tonynelson at georgeanelson.com>
      '      is no more.             <http://www.georgeanelson.com/>