Re: make: avoid breaking backwards compatability

[David Holland <dholland-tech%netbsd.org@localhost>]

On Mon, Sep 01, 2014 at 12:15:58PM -0700, Simon Gerraty wrote:
 > The recent changes to suffix white-space handling are cool,
 > but they break a number of existing makefiles.
 > [...]
 > 
 > A target that has a \ at the end of the last line of its script now breaks.
 > Such targets were valid until recently and the practice is not uncommonn
 > since it reduces noise in diff's when additional commands added
 > (eg. great log sed commands)

Oh please, that's a bug in the makefile. What does an escaped nothing
mean? If you're so concerned about the diffs, stick a blank line at
the end.

And yeah, that means changing some makefiles; if you've been doing
something so obviously invalid you need to expect that. Sorry.

 > The other case I hit was where SRCS contains sub/file.c 
 > sometimes adding the dir where sub/file.c lives to .PATH isn't an option
 > because of other files present that are not the desired ones.
 > 
 > Until recently
 > 
 > foo.o:       subdir/foo.c
 > 
 > worked, but now even after subdir/foo.c has been correctly found
 > a search for foo.c is made and overrides the previous result - which can
 > cause the wrong foo.c to be compiled.
 > [...]

Why add a knob for this? It is just a bug. Unless there is some part
of the recent set of discussions and changes that I've missed (which
may be the case, and I haven't read all of it in detail either because
it's large and I've been waiting for an uninterrupted chunk of time) I
don't see how this pertains to any of the cited POSIX issues.

If you understand where the wrong behavior comes from, please fix it.
I do not, even after looking through your patch, and the patch itself
seems questionable as it seems to add a special case rather than
dealing with the issue in a systematic way.

Since .PATH is only supposed to be used for files that are not found
in the current directory, and in the case you care about subdir/foo.c
is found in (or at least relative to) the current directory, it seems
that there should not need to be special-case handling to prevent
searching for it.

Even if the recent batch of changes is wrong, please don't add
special-case bodges. I'd rather revert the recent changes (that has
already been seeming like a good idea) and rework them.

Make has too many special-case bodges already.

-- 
David A. Holland
dholland%netbsd.org@localhost