On Mon, Jul 21, 2014 at 11:07:18AM +0200, Pietro Cerutti wrote: > On 2014-Jul-18, 18:47, Bryan Drewery wrote: > > On 2014-07-10 08:08, Pietro Cerutti wrote: > > > Author: gahr (ports committer) > > > Date: Thu Jul 10 13:08:51 2014 > > > New Revision: 268494 > > > URL: http://svnweb.freebsd.org/changeset/base/268494 > > > > > > Log: > > > Implement Short/Small String Optimization in SBUF(9) and change > > > lengths and > > > positions in the API from ssize_t and int to size_t. > > > > > > CR: D388 > > > Approved by: des, bapt > > > > > > Modified: > > > head/lib/libsbuf/Makefile > > > head/sys/kern/subr_sbuf.c > > > head/sys/sys/sbuf.h > > > > > > Modified: head/lib/libsbuf/Makefile > > > ============================================================================== > > > --- head/lib/libsbuf/Makefile Thu Jul 10 13:04:52 2014 > > > (r268493) > > > +++ head/lib/libsbuf/Makefile Thu Jul 10 13:08:51 2014 > > > (r268494) > > > @@ -4,7 +4,7 @@ LIB= sbuf > > > SHLIBDIR?= /lib > > > SRCS= subr_sbuf.c > > > > > > -SHLIB_MAJOR = 6 > > > +SHLIB_MAJOR = 7 > > > > > > SYMBOL_MAPS= ${.CURDIR}/Symbol.map > > > VERSION_DEF= ${.CURDIR}/Version.def > > > > > > > We have symbol versioning in libsbuf. Shouldn't need to bump the > > version. Just need to add a compat function for everything. It would be > > a lot of work, but bumping for ssize_t->size_t seems gratuitous to me. > > The bump was due mainly to the addition of the s_static_buf field to the > sbuf struct.
So, how this invalidates the point of not breaking the ABI ? We do not bump .so version for versioned libraries. For all symbols (functions) which are affected by the change, the compat shims in the old namespace must be provided.
pgpmttjOL06kP.pgp
Description: PGP signature