Re: interface to make?

[Glenn Fowler]

this web paper from 1994

        http://www.research.att.com/~gsf/mam/

describes the MAM (make abstract machine) language for instrumenting
make for static, runtime, and regression analysis

MAM is strictly readonly; there is no control channel back to make
MAM output is written to a file and there are no other signals or events

the paper provides details, with examples, and also a link to a
patch that adds MAM to make-3.81beta1

a popular MAM application renders makefile dependency graphs
check out the AST nmake dependency graph for make-3.81beta1

        http://www.research.att.com/~gsf/mam/gmake-nmake.ps.gz

most MAM output is the result of "walking the graph":
        pattern metarule application,
        variable assignments (listed on first use)
        per-target variable assignments,
        action success or failure,
        concurrent action execution

although MAM is similar to -d output, MAM syntax and semantics
provide a stable base for makefile analysis

-- Glenn Fowler -- AT&T Labs Research, Florham Park NJ --

%% maman yonatan <address@hidden> writes:

  my> is there any mechanism for subscribing make events,
  my> like new rule was found, new variable was set, a
  my> pattern stem was matched etc. ?

> Paul Smith wrote:

> I don't quite understand what you're asking.  To whom would make send
> any events, signals, etc.?  There is no "make library" that could be
> linked with another program, if that's what you're asking.
> 
> Also, note that make proceeds in two distinct steps: in the first step
> it reads in all the makefiles and constructs an internal graph of all
> the dependency relationships, etc.  In the second step it walks the
> graph and actually invokes the rules.  So, your events such as setting a
> variable and "finding" a new rule, which would happen in the first step,
> are very different from matching a pattern, which would happen in the
> second step.