[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: SONAME of librpm



On Tue, Jul 03, 2001 at 04:33:30PM -0400, Ian Peters wrote:
> On Tue, Jul 03, 2001 at 03:43:13PM -0400, Jeff Johnson wrote:
> > On Tue, Jul 03, 2001 at 04:34:45PM -0300, Ruda Moura wrote:
> > > Why is librpm still named librpm.so.0 in rpm 4.0.x even having a
> > > different API from rpm 3.0.x?
> > > 
> > 
> > Because the ABI is largely (if not entirely) unchanged, and the change to
> > the API is mostly (if not entirely) unchanged.
> 
> What are you talking about?
> 
> All of the db iterator stuff, the random API changes you sneak in
> because you think no one is using them, the rpmProblem changes you
> snuck in between 4.0 and 4.0.2...
> 

The random *documented* API/ABI changes that I sneak in, please, not that
it matters much.

> > Because most of the breakage is due to mobving a big hunk of the ABI into
> > a new library, and changing the soname has nothing to do with that. Linking
> > libraries might address the issue, but there are reasons not to do that
> > included in the libtool info page.
> 
> Moving API to a different library means the API exposed by the first
> library is different (in this case smaller).  So you bump the .so
> number.
> 

Not entirely my call. Nor is it my call for rpm-4.0.3, the soname change is
deemed "too large" for a minor release, said release "4.0.3" not chosen by me.
So, rather than change the rpm version numbet that I've been distributing
for months, I back off changing the soname Yet Again.

> > Because.
> 
> It terrifies me that a fairly crucial component of many systems is
> maintained in such an ad hoc way.
> 

I'm shocked, simply shocked, as well. I also recommend and encourage static
linkage (as you know).

73 de Jeff

-- 
Jeff Johnson	ARS N3NPQ
jbj@jbj.org	(jbj@redhat.com)
Chapel Hill, NC





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []