keep generated code in Samba4?

[Andrew Bartlett]

On Thu, 2003-11-20 at 21:19, Simo Sorce wrote:
> On Thu, 2003-11-20 at 07:00, tridge at samba.org wrote:
> > We now have around 50k lines of generated RPC code in Samba4, and its
> > growing fast. It might make sense to not put this code in CVS, and
> > instead rely on all developers having the pidl IDL compiler installed,
> > or possibly even including pidl in the Samba4 cvs tree (pidl is much
> > smaller than the code it generates).

I strongly agree with keeping such trivially generated files out of
CVS.  

> > Anyone have any special reason one way or the other? Up to now we have
> > not relied on perl to build Samba, but perhaps that should change. For
> > example a perl mkproto would probably be much faster and neater than
> > our current awk one, and perl could be useful in some other parts of
> > the build.
> 
> Currently I had the same issue with HEAD, I added the genstruct stuff
> modified for samba, but as HEAD never happened to be perl dependent I
> also made a new target in Makefile so that the developer can build the
> needed headers and commit them. 

I think a machine that can't install perl is a pretty poor machine
indeed.  I would suggest perl is probably more ported than Samba is,
these days.

> Not putting the generated code into the samba4 cvs is wise, as what we
> must rely on is _only_ the idl, or you would risk some commit not in
> line with the idls.
> But if you do that you should also include pidl in the samba4 cvs unless
> it becomes a separate package available for most distributions.

I would suggest keeping pidl where it is - if you are brave enough to
play with samba4, finding another samba.org CVS module isn't going to be
an issue.  

That said, there is one reason to move pidl a matter of 'configuration
managment'.  That is, do we really want to be debugging bugs that are
related to a build of samba4 being built with an old pidl?  

Likewise, if one was to test older versions, getting a matching pidl
version would be important.

> > We might also decide to use an interim method, where we do not put the
> > generated code in CVS, but we generate the code on samba.org before
> > each release, so releases have the full generated code but developers
> > who use cvs need to have pidl.
> 
> This is "the classic" way adopted in samba, but this time the amount of
> code is so mush that putting pidl into samba4 is preferable imho.

We need to generate the code on samba.org only if pidl is seperate, and
possibly for the build farm too.  I don't see these as being an issue.

Andrew Bartlett

-- 
Andrew Bartlett                                 abartlet at pcug.org.au
Manager, Authentication Subsystems, Samba Team  abartlet at samba.org
Student Network Administrator, Hawker College   abartlet at hawkerc.net
http://samba.org     http://build.samba.org     http://hawkerc.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.samba.org/archive/samba-technical/attachments/20031120/ab6631ce/attachment.bin