I am not quite the autoconf expert you are looking for, but the basic
plan for distcheck is to unpack the distfile, and then to do an objdir
build and then make check, to validate that the distfile was produced
correctly.

I don't think "steps are not in same order" is a fair complaint because
to be pedantic make doesn't have a spec for an order.  It is merely
obligated to respect the expressed dependencies, and various things may
be a little different.  So the real question is if the order is wrong
and how you can argue that.

I would ask:

  If make is building something before something it needs is built, how
  are you expressing that in the makefile?

  What if you, on a machine with lots of CPUs, do "make -j32" in the
  original build?  is that 100% reliable?

  What if you  take the distfile to another machine, unpack it, mkdir
  BUILD, cd BUILD, ../configure, make, make check?  Does that work?

Reply via email to