Re: Order of include and variable expansion

[Paul Millar]

Hi Dave,

Thanks for replying.

On Sunday 11 November 2007 14:20:04 Dave Korn wrote:
> On 10 November 2007 17:52, Paul Millar wrote:
> > I've a makefile of medium complexity
> >
> > I have a work-around for this, but I was wondering if:
> >   a. is this a bug or a known problem?
> >
> >   b. whether there's anyway of indicating the order in which phase 1
> > activity should take place (e.g. do the include first, delay expanding
> > the prerequisites, that sort of thing).
> >
> > I can post the makefile somewhere, if that helps.
>
>   What would really help would be a cut-down trivial testcase, with just
> enough in it to show the required dependencies and the autogeneration. 
> It's pretty hard to know what's going on without seeing an example, but the
> first thing I'd look at in such a testcase would be whether you're running
> into problems from the use of ':=' immediate evaluation versus '=' deferred
> evaluation.

I've attached a small test-case (it's very small, so I hope emailing it to the 
list is OK).  This test-case replicates the problem without using include, so 
this may not have been the root-cause.

There are four parts to the test-case:

        First, run the makefile as-is.  This should demonstrate correct 
behaviour.

        Second, comment the first TEST_LIST definition and uncomment the second 
TEST_LIST definition.  The output should be different and, if run 
with --warn-undefined-variables, you should get a warning.

        Third, uncomment the PHONY definition.  You should find the problem 
persists.

        Forth, uncomment the initial TEST_LIST definition and comment the 
second 
definition.  You should *still* see the problem persist.

Only when the PHONY is commented again does the makefile work.

I'm using make v3.81 on Debian sid.

HTH,

Paul.

Makefile
Description: Text Data