[Nix-dev] Re: [PATCH 1/2] t/scons

[Peter Simons]

Hi Marc,

 >> this patch changes SCons' documented behavior. The authors of the tool
 >> have made the conscious choice to ignore the user's shell environment.
 >
 > Try thinking about it in a different way: In the past mathematicians
 > learned about negative numbers. New case - new ways to think about
 > existing laws.

it's not obvious to me what conclusions we are supposed to draw from that
analogy. The SCons manual states:

 | scons does not automatically propagate the external environment used
 | to execute scons to the commands used to build target files.

Your patch, however, modifies SCons to propagate the external environment to
the build environment. Consequently, that patch changes the specified
behavior of the tool. If we apply that patch, then it's going to come as a
major surprise to SCons users, who expect it to behave differently. On these
grounds, it is unwise to modify the derivation called "scons" in the way you
propose. Instead, that patch should be applied in a different expression,
say "scons-path-auto-import", or whatever, so that nixpkgs users have the
chance to distinguish pristine SCons from a patched one.

Frankly, I don't understand why you resist that kind of non-intrusive
solution so much. What is wrong with a separate expression? You can override
the 'scons' attribute locally on your machine as you please! Why is that not
good enough? Why do you feel you must change the default behavior for
everyone else too?

Take care,
Peter