Re: targets/SECONDARY TEST#9 fails sometimes

[Tim Newsome]

Does anybody know what the right solution to this problem is? I'm happy to implement/test either one, but I don't want to spend my time working on a solution that won't be accepted upstream.

Tim

On Tue, Nov 22, 2011 at 3:16 PM, Tim Newsome <address@hidden> wrote:

TEST #9 in targets/SECONDARY sometimes fails when two commands execute in a different order than expected. As far as I can tell make is doing the right thing.

# TEST #9 -- Savannah bug #15919

# The original fix for this bug caused a new bug, shown here.

touch(qw(1.a 2.a));

run_make_test('

%.c : %.b ; cp $< $@

%.b : %.a ; cp $< $@

all : 1.c 2.c', '-rR -j',

'cp 1.a 1.b

cp 2.a 2.b

cp 1.b 1.c

cp 2.b 2.c

rm 1.b 2.b');

unlink(qw(1.a 2.a 1.c 2.c));

The expected order of cp statements is above, but sometimes I see the following:

cp 1.a 1.b

cp 2.a 2.b

cp 2.b 2.c

cp 1.b 1.c

rm 1.b 2.b

I can think of two ways to fix this:

1. Make 2.c depend on 1.c.

2. Change the verification code to accept both orders.

What is the right solution?

Tim