Using public headers to write ABI vscripts

[Andrew Bartlett]

On Fri, 2012-08-31 at 09:54 +0300, Alexander Bokovoy wrote:
> On Fri, Aug 31, 2012 at 9:46 AM, Andrew Bartlett <abartlet at samba.org> wrote:
> > On Fri, 2012-08-31 at 09:38 +0300, Alexander Bokovoy wrote:
> >> On Fri, Aug 31, 2012 at 9:06 AM, Andrew Bartlett <abartlet at samba.org> wrote:
> >> > On Fri, 2012-08-31 at 08:31 +0300, Alexander Bokovoy wrote:
> >> >> Hi Andrew,
> >> >>
> >> >> On Wed, Aug 29, 2012 at 7:54 AM, Andrew Bartlett <abartlet at samba.org> wrote:
> >> >> > I've been working on a patch to use the mksyms.awk approach to get a
> >> >> > list of public headers and symbols from our public headers files, and
> >> >> > use that instead of just the regular-expression abi_match.
> >> >> >
> >> >> > https://git.samba.org/?p=abartlet/samba.git/.git;a=shortlog;h=refs/heads/abi-public-headers
> >> >> >
> >> >> > It would be great to get this in to 4.0, but I'm having some waf
> >> >> > challenges I'll need to get past.  The challenge relates to the grouping
> >> >> > from the build ordering, because we need a vscript for libreplace (very
> >> >> > early) but vscripts can depend on pild-generated headers.  We may need
> >> >> > to rely on pure dependencies rather than also the groups.
> >> >> >
> >> >> > This came up because I had to bump the ldb ABI (it needed a bump for the
> >> >> > behaviour change, but it spiked my interest) when adding a private,
> >> >> > internal symbol.
> >> >> I'm also working in this area though my needs are bit different at the moment.
> >> >>
> >> >> https://git.samba.org/?p=ab/samba.git/.git;a=shortlog;h=refs/heads/stableabi
> >> >> has number of patches that attempt to introduce API namespaces.
> >> >>
> >> >> I need stable mapping between API namespace and version node name
> >> >> associated with it that spans across multiple releases (instead of
> >> >> SAMBA_<VERSION> default symbol changing every release without actual
> >> >> semantical and ABI change) for certain important private libraries.
> >> >
> >> > I'm a bit confused.  In what sense is a library that is being linked
> >> > from outside our build process still private?
> >> libpdb is required to link against if you are passdb module. FreeIPA
> >> provides its ipasam module and it needs to link against libpdb to be
> >> useful in smbd.It is consumed purely by smbd and thus still considered
> >> private but the source code lives outside samba source tree and thus
> >> public.
> >
> > I really don't like this half-public-half-private thing.  Either this is
> > a public library, or a private library, it shouldn't be both.  We
> > certainly shouldn't be installing public headers for a private library
> > like we currently do.
> >
> > Certainly an external package should *never* be linking directly into
> > our private library folder.
> >
> > We have other libraries that we do not make strong promises regarding
> > ABIs for, but which we must provide external access (samba-util comes to
> > mind).  These we still declare as public.
> I'm fine if we would declare libpdb and libsmbldap as 'public'.
> Cleaning them up and splitting truly internal code is something I can
> do.
> 
> >> > I don't like the idea of public users of our private libraries.  If we
> >> > wish to declare a library open for public use, shouldn't it just be
> >> > given a version number and be declared as public?
> >> libpdb includes all sorts of things, not only API that is used by the
> >> modules. For example, all statically compiled pdb modules are compiled
> >> into libpdb.
> >
> > Then we need to rework things so that we provide a public library that
> > is externally useful and acceptable, containing only the bare minimum
> > that writing an external passdb module would need.  That may mean having
> > a different private library that handles the registration (and therefore
> > potentially static linking) of the passdb modules.
> I did try that in first revision of the patchset. I quickly got into
> circular dependencies hell, unfortunately, as all of pdb modules,
> including the statically linked ones, require the same base set of
> registration functions that embed statically linked modules (in
> lazy_initialize_passdb()).

Getting rid of lazy_initialize_passdb() seems like a good way to break
that.  It makes little sense anyway, we already have
initialize_password_db(), which must be called anyway to get the
pdb_tevent_cxt initialised for pdb_ldap/smbldap idle events.  

Andrew Bartlett
-- 
Andrew Bartlett                                http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org