[src/trunk]: src/usr.bin/make/unit-tests tests/make: make test job-output-nul...

[rillig <rillig%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/9a216ab66e2c
branches:  trunk
changeset: 984337:9a216ab66e2c
user:      rillig <rillig%NetBSD.org@localhost>
date:      Sat Jul 03 22:39:40 2021 +0000

description:
tests/make: make test job-output-null more portable

Previously, the test dependend on implementation details of the system's
printf command.

Thank you sjg for the detailed analysis on macOS, FreeBSD and Linux.

diffstat:

 usr.bin/make/unit-tests/job-output-null.exp |   2 +-
 usr.bin/make/unit-tests/job-output-null.mk  |  21 +++++++++++++++++----
 2 files changed, 18 insertions(+), 5 deletions(-)

diffs (60 lines):

diff -r 4fce1127116a -r 9a216ab66e2c usr.bin/make/unit-tests/job-output-null.exp
--- a/usr.bin/make/unit-tests/job-output-null.exp       Sat Jul 03 21:27:48 2021 +0000
+++ b/usr.bin/make/unit-tests/job-output-null.exp       Sat Jul 03 22:39:40 2021 +0000
@@ -1,4 +1,4 @@
 hello
 hello
-hello world without   newline, hello world without   newline, hello world without   newline.
+hello without   newline, hello without   newline.
 exit status 0
diff -r 4fce1127116a -r 9a216ab66e2c usr.bin/make/unit-tests/job-output-null.mk
--- a/usr.bin/make/unit-tests/job-output-null.mk        Sat Jul 03 21:27:48 2021 +0000
+++ b/usr.bin/make/unit-tests/job-output-null.mk        Sat Jul 03 22:39:40 2021 +0000
@@ -1,4 +1,4 @@
-# $NetBSD: job-output-null.mk,v 1.1 2021/04/15 19:02:29 rillig Exp $
+# $NetBSD: job-output-null.mk,v 1.2 2021/07/03 22:39:40 rillig Exp $
 #
 # Test how null bytes in the output of a command are handled.  Make processes
 # them using null-terminated strings, which may cut off some of the output.
@@ -7,20 +7,33 @@
 # inconsistently.  It's an edge case though since typically the child
 # processes output text.
 
+# Note: The printf commands used in this test must only use a single format
+# string, without parameters.  This is because it is implementation-dependent
+# how many times the command 'printf "fmt%s" "" "" ""' calls write(2).
+#
+#      NetBSD /bin/sh          1 x write("fmtfmtfmt")
+#      Dash                    1 x write("fmtfmtfmt")
+#      NetBSD /bin/ksh         3 x write("fmt") (via /bin/printf)
+#      Bash 5                  3 x write("fmt")
+#
+# In the latter case the output may arrive in parts, which in this test makes
+# a crucial difference since the outcome of the test depends on whether there
+# is a '\n' in each of the blocks from the output.
+
 .MAKEFLAGS: -j1                # force jobs mode
 
 all: .PHONY
        # The null byte from the command output is kept as-is.
        # See CollectOutput, which looks like it intended to replace these
        # null bytes with simple spaces.
-       @printf 'hello\0world%s\n' ''
+       @printf 'hello\0world\n'
 
        # Give the parent process a chance to see the above output, but not
        # yet the output from the next printf command.
        @sleep 1
 
        # All null bytes from the command output are kept as-is.
-       @printf 'hello\0world%s\n' '' '' '' '' '' ''
+       @printf 'hello\0world\n''hello\0world\n''hello\0world\n'
 
        @sleep 1
 
@@ -29,4 +42,4 @@
        #
        # The three null bytes in a row test whether this output is
        # compressed to a single space like in DebugFailedTarget.  It isn't.
-       @printf 'hello\0world\0without\0\0\0newline%s' ', ' ', ' '.'
+       @printf 'hello\0without\0\0\0newline, hello\0without\0\0\0newline.'