On Fri, 2007-04-20 at 13:13 +0200, Patrice Dumas wrote:
On Fri, Apr 20, 2007 at 01:06:51PM +0200, Ralf Corsepius wrote:
The guidelines intention is to recommend "foo-static".
Ok, so what about rpmlint warnings? Ignore them or bugzilla rpmlint?
bugzilla it against rpmlint, if you like to.
I don't know where this rpmlint warning stems from.
If I was to decide, I would reject any static library unless a pressing need of requiring a static lib can be demonstrated (!).
There was a very long thread about that issue and I'd like not to repeat it all over again. In summary, there are cases, like numerical models where all the reason that are for dynamic libraries don't make sense (use new version, security, dlopening, name resolution...) while compiling statically helps being able to move programs to other hosts/linux distros/ and so on. The other solution advocated instead of static linking (distributing shared libs along with the program and doing a wrapper that setes things up) is much less practical.
Another reason may be the efficiency, as as demonstrated by Enrico numbers on tiny daemons, linking statically may lead to much more efficient executables.
What some people call efficiency, I call rending a distro unmaintainable and pimping "Linux" - To me, it's not much different from pimping a car by installing an oxygen bottle to "make it faster".
(But there are cases when user should be able to link against static libs, a prominent case -- my case -- being numerical models).
You know my opinion on this argument of yours: You are abusing Linux.
Not at all. I have specific needs.
Well, your needs, i.e. "cross-distro binaries", are far from being exotic. Many people before you went into trap you still seem to be trapped into, before you, so be it :-)
Ralf