regress/infra-unittests: demonstrate wrong SUBST no-op failure

Discovered by Juraj in mail/policyd-weight.

https://mail-index.netbsd.org/pkgsrc-changes/2020/06/06/msg215480.html
This commit is contained in:
rillig 2020-06-06 13:00:52 +00:00
parent 3d4dde697c
commit 4eaad76616

View file

@ -1,5 +1,5 @@
#! /bin/sh
# $NetBSD: subst.sh,v 1.43 2020/05/19 05:14:18 rillig Exp $
# $NetBSD: subst.sh,v 1.44 2020/06/06 13:00:52 rillig Exp $
#
# Tests for mk/subst.mk.
#
@ -1634,3 +1634,48 @@ if test_case_begin 'multiple sed commands with semicolon'; then
test_case_end
fi
if test_case_begin 'several substitution, only one applies to file'; then
# https://mail-index.netbsd.org/pkgsrc-changes/2020/06/06/msg215480.html
create_file 'testcase.mk' <<-EOF
SUBST_CLASSES+= id
SUBST_FILES.id= first second
SUBST_SED.id= -e 's,first,first-modified,'
SUBST_SED.id+= -e 's,second,second,'
SUBST_NOOP_OK.id= no
.include "prepare-subst.mk"
.include "mk/subst.mk"
EOF
create_file_lines 'first' 'first'
create_file_lines 'second' 'second'
run_bmake 'testcase.mk' 'subst-id' 1> "$tmpdir/output" 2>&1 \
&& exitcode=0 || exitcode=$?
# FIXME: This is not the intended behavior.
# Each of the files contains at least one of the sed patterns,
# therefore the substitutions _could_ have an effect, depending
# on the pkgsrc configuration.
#
# Instead of testing whether the sed command is an identity
# substitution, the original idea was to find files that contain
# _none_ of the patterns.
#
# The proper fix might be as simple as replacing the "all" with an
# "any" in mk/scripts/subst-identity.awk. This would also nicely
# remove the special handling of an empty command line.
assert_that "$tmpdir/output" --file-is-lines \
'=> Substituting "id" in first second' \
'warning: [subst.mk:id] Nothing changed in "second".' \
'fail: [subst.mk:id] The filename pattern "second" has no effect.' \
'*** Error code 1' \
'' \
'Stop.' \
"$make: stopped in $PWD"
test_case_end
fi