Re: [PATCH] Document AM_MISSING_PROG.

[Benoit Sigoure]

On Nov 5, 2007, at 10:09 PM, Ralf Wildenhues wrote:

> > On Nov 4, 2007, at 1:49 AM, Ralf Wildenhues wrote:
> > > This is fine with me, given that we document AM_MISSING_PROG in
> > > Automake.  That is not completely trivial, though: typically, the
> > > "missing" script needs adjustment for tools, making it un-generic.
> >
> > I did not quite understand what you meant here, Ralf, so I didn't  
> > do anything
> > special.  Are you refering to the --run option of `missing'?
>
> For example.  For another example, as a user I'd ask myself: what are
> the semantics of "some arbitrary program" being run my `missing'?
> (If we don't define any, we risk that the user just assumes whatever
> currently happens; that prevents us from improving them later.)
>
> Note that `missing' only helps really well for programs that it knows,
> where it can emulate updating the target file so that the rule isn't
> rerun every time.  This limits its usefulness for arbitrary programs.
>
> I don't know how this fact can be documented well, or FWIW, this
> limitation removed (plugins in shell?  tell the user to patch
> `missing'?)

OK I see what you mean now.  Actually, I always used `missing' as a  
mean to `print a nice message to the user if they try to do something  
that requires a missing tool', not as a replacement program to  
emulate  functionalities of missing programs.  Heh.

> Note also that there is existing documentation for `missing'.  This  
> new
> part should interact with it (cross-reference if not close in  
> location).

Hmm yeah.

> > By the way, the TODO file seems to have many rusted entries that  
> > have been
> > lying there for too long or completed since a long time already.
>
> Yes.  If you could provide a proper patch, that would help.  But  
> please
> make sure that the stuff you remove really has been fixed properly,
> that's the big part of the work.  Some of the entries are rather
> mysterious and some are difficult to understand.

Yes, that's a tough task.

> [...]
> > +cat > Makefile.am << 'EOF'
> > +all-local:
> > +       $(MYSHELL) -c 'echo hi >foo'
> > +EOF
> > +
> > +cat >> configure.in << 'EOF'
> > +AM_MISSING_PROG(MYSHELL, sh)
> > +AC_OUTPUT
> > +EOF
> > +
> > +$ACLOCAL
> > +$AUTOCONF
> > +$AUTOMAKE --add-missing
> > +
> > +./configure --prefix "`pwd`"
> > +$MAKE
> > +
> > +echo hi >expected
> > +
> > +cmp foo expected
>
> Would be good to also have a test with a program that does not exist,
> i.e., a test that exposes all aspects of `missing's functionality,
> OTOH, tests/missing*.test covers most of that already, so this is ok
> I guess.

That's what I thought, do I need to write more tests?  I find the  
testsuite slow enough despite my 2Ghz Core2Duo ;o  (It'll be better  
with make -j2, once we include the parallel test feature :D)

I'll try to post a revised version of that patch in a couple of days.

--
Benoit Sigoure aka Tsuna
EPITA Research and Development Laboratory

PGP.sig
Description: This is a digitally signed message part