[Fwd:[ksh93-integration-discuss]ksh93-integrationpre-reviewround"two"(webrev2007-05-14)]

[Roland Mainz]

Peter Memishian wrote:
> 
>  > > lib/lvm, where all of the libraries sharing a given Makefile "architecture"
>  > > would exist below lib/ast (e.g., lib/ast/libast, lib/ast/libpp, ...), and
>  > > then the two Makefiles could live under lib/ast/Makefile.ast*.  But I'm
>  > > guessing you considered that approach and had an issue with it -- so the
>  > > next best thing would be to have those Makefiles live directly in lib.
>  >
>  > The issue is that if OpenSolaris develops an add-on library (which AFAIK
>  > will happen soon for the "open"/"close"/"poll"/etc. builtins (e.g.
>  > compile the code and do a "builtin -f libshell.so poll ; poll --man #
>  > :-) ) and "procsh") and contributes this to upstream - will this library
>  > live under usr/src/lib/ast/, usr/src/lib/ or elsewhere ? AFAIK this is
>  > quickly becoming messy because there are no clear lines if libshell&co.
>  > get tightly integrated with some consumers within OS/Net and still
>  > sync'ed with upstream and/or other projects...
> 
> If it needed to use the Makefile.ast* logic, it would live under lib/ast.
> Otherwise, it wouldn't.

Uhm... IMHO that's wrong. Subdirs should be project-specific (if there
is a strong need (and good justification) for such a thing (which I
don't see for libshell&co. - remeber both AST and Solaris's base code
(System V) share the same origin)) and not based on the underlying APIs
used by the code.

[snip]
>  > and ROOTDEMODIRBASE depends on the demo source sitting in
>  > "common/".
> 
> Could it use $(SRCDIR) directly (and then one could conditionally override
> $(SRCDIR) for $(ROOTDEMOFILES) if necessary?)

SRCDIR is not set in this level. We're talking about
usr/src/lib/lib*/Makefile, not usr/src/lib/lib*/$(TRANSMACH)/Makefile
...

>  > 2. Is there any recommendation what I should do when the consumer
>  > Makefiles of Makefile.lib do not have any demo files ? If ROOTDEMODIRS,
>  > ROOTDEMOFILES, ROOTDEMODIRBASE aren't defined AFAIK  all hell will break
>  > loose - which means we need some kind of "dummy" values.
> 
> Doesn't seem like $(ROOTDEMOFILES) being empty would be a problem.  For
> the other two, the convention is to use __nonexistent_directory__, as
> we do with HDRDIR (and others).

Ok...

... patch is done, I am waiting for a quick $ (cd usr/src ; dmake
install) # to complete (and in the meantime I fetch some black tea...)
...

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) roland.mainz at nrubsig.org
  \__\/\/__/  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
  /O /==\ O\  TEL +49 641 7950090
 (;O/ \/ \O;)