Subject: bin/14544: [rkb] make search path oddity causes unexpected failures
To: None <firstname.lastname@example.org>
From: None <email@example.com>
Date: 11/11/2001 16:14:32
>Synopsis: make search path oddity causes unexpected failures
>Arrival-Date: Sun Nov 11 13:15:00 PST 2001
>Release: NetBSD 1.5Y, complete build from cvs of Nov 6th, 2001
Not today, sorry.
System: NetBSD cyclops 1.5Y NetBSD 1.5Y (CYCLOPS) #3: Tue Nov 6 08:30:46 EST 2001 rafal@cyclops:/extra/src-current/sys/arch/i386/compile/CYCLOPS i386
Architecture and Machine are i386, but I don't think this is an MD bug.
Assume we have a directory tree like the following:
And assume that topdir/foo has a (already make) target file `bar'.
A make from topdir/foo attempting to make `foo/bar' will fail to
make the `bar' target in `topdir/foo/foo/' as make incorrectly
finds `topdir/foo/bar', claims it's up to date and quits.
Attached below is a uuencoded .tar.gz file that includes a sample
Untar this file, cd to foo/ and type `make'.
It's expected that `make' will copy `foo/foo/Makefile.in' to
`foo/foo/Makefile' via the foo/Makefile target.
However, with a make built from Nov 6th sources, `make' finds
`foo/Makefile', declares it up-to-date, and does *not* run the
rules to make `foo/foo/Makefile'.
This bites the xsrc build bigtime, as the Makefile-generating
machinery uses similar rules to make Makefiles out of Imakefiles.
Some investigation by Alasdair Baird have shown that sources of
October 31, 2001 are OK.. See the below URL.
More details on how this affects xsrc can be found in the mailing
list archives at:
begin 644 make-failure.tgz