Module Name: src
Committed By: rillig
Date: Fri Aug 28 17:15:04 UTC 2020
Modified Files:
src/usr.bin/make: make.1
src/usr.bin/make/unit-tests: varmod-subst-regex.exp
varmod-subst-regex.mk
Log Message:
make(1): fix manual page about the modifiers of the :C modifier
To generate a diff of this commit:
cvs rdiff -u -r1.288 -r1.289 src/usr.bin/make/make.1
cvs rdiff -u -r1.2 -r1.3 src/usr.bin/make/unit-tests/varmod-subst-regex.exp \
src/usr.bin/make/unit-tests/varmod-subst-regex.mk
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
Modified files:
Index: src/usr.bin/make/make.1
diff -u src/usr.bin/make/make.1:1.288 src/usr.bin/make/make.1:1.289
--- src/usr.bin/make/make.1:1.288 Sat Aug 22 19:30:58 2020
+++ src/usr.bin/make/make.1 Fri Aug 28 17:15:04 2020
@@ -1,4 +1,4 @@
-.\" $NetBSD: make.1,v 1.288 2020/08/22 19:30:58 sjg Exp $
+.\" $NetBSD: make.1,v 1.289 2020/08/28 17:15:04 rillig Exp $
.\"
.\" Copyright (c) 1990, 1993
.\" The Regents of the University of California. All rights reserved.
@@ -29,7 +29,7 @@
.\"
.\" from: @(#)make.1 8.4 (Berkeley) 3/19/94
.\"
-.Dd August 22, 2020
+.Dd August 28, 2020
.Dt MAKE 1
.Os
.Sh NAME
@@ -1310,15 +1310,15 @@ in each word of the variable's value, re
.Ar new_string .
If a
.Ql g
-is appended to the last slash of the pattern, all occurrences
+is appended to the last delimiter of the pattern, all occurrences
in each word are replaced.
If a
.Ql 1
-is appended to the last slash of the pattern, only the first occurrence
+is appended to the last delimiter of the pattern, only the first occurrence
is affected.
If a
.Ql W
-is appended to the last slash of the pattern,
+is appended to the last delimiter of the pattern,
then the value is treated as a single word
(possibly containing embedded white space).
If
@@ -1387,13 +1387,6 @@ as occur in the word or words it is foun
.Ql W
modifier causes the value to be treated as a single word
(possibly containing embedded white space).
-Note that
-.Ql 1
-and
-.Ql g
-are orthogonal; the former specifies whether multiple words are
-potentially affected, the latter whether multiple substitutions can
-potentially occur within each affected word.
.Pp
As for the
.Cm \&:S
Index: src/usr.bin/make/unit-tests/varmod-subst-regex.exp
diff -u src/usr.bin/make/unit-tests/varmod-subst-regex.exp:1.2 src/usr.bin/make/unit-tests/varmod-subst-regex.exp:1.3
--- src/usr.bin/make/unit-tests/varmod-subst-regex.exp:1.2 Sun Aug 16 12:30:45 2020
+++ src/usr.bin/make/unit-tests/varmod-subst-regex.exp Fri Aug 28 17:15:04 2020
@@ -1,12 +1,5 @@
-mod-regex:
-:a b b c:
-:a b b c:
-: b c:
make: Regex compilation error: (details omitted)
-:C,word,____,:Q}:
-:a c:
-:x__ 3 x__ 3:
-mod-regex-limits:00-ok:1 2323 45456
+mod-regex-compile-error: C,word,____,:Q}.
make: No subexpression \1
make: No subexpression \1
make: No subexpression \1
Index: src/usr.bin/make/unit-tests/varmod-subst-regex.mk
diff -u src/usr.bin/make/unit-tests/varmod-subst-regex.mk:1.2 src/usr.bin/make/unit-tests/varmod-subst-regex.mk:1.3
--- src/usr.bin/make/unit-tests/varmod-subst-regex.mk:1.2 Sun Aug 16 12:30:45 2020
+++ src/usr.bin/make/unit-tests/varmod-subst-regex.mk Fri Aug 28 17:15:04 2020
@@ -1,22 +1,79 @@
-# $NetBSD: varmod-subst-regex.mk,v 1.2 2020/08/16 12:30:45 rillig Exp $
+# $NetBSD: varmod-subst-regex.mk,v 1.3 2020/08/28 17:15:04 rillig Exp $
#
# Tests for the :C,from,to, variable modifier.
-all: mod-regex
+all: mod-regex-compile-error
all: mod-regex-limits
all: mod-regex-errors
-mod-regex:
- @echo $@:
- @echo :${:Ua b b c:C,a b,,:Q}:
- @echo :${:Ua b b c:C,a b,,1:Q}:
- @echo :${:Ua b b c:C,a b,,W:Q}:
- @echo :${:Uword1 word2:C,****,____,g:C,word,____,:Q}:
- @echo :${:Ua b b c:C,b,,g:Q}:
- @echo :${:U1 2 3 1 2 3:C,1 2,___,Wg:C,_,x,:Q}:
+# The variable expression expands to 4 words. Of these words, none matches
+# the regular expression "a b" since these words don't contain any
+# whitespace.
+.if ${:Ua b b c:C,a b,,} != "a b b c"
+.error
+.endif
+# Using the '1' modifier does not change anything. The '1' modifier just
+# means to apply at most 1 replacement in the whole variable expression.
+.if ${:Ua b b c:C,a b,,1} != "a b b c"
+.error
+.endif
+
+# The 'W' modifier treats the whole variable value as a single big word,
+# containing whitespace. This big word matches the regular expression,
+# therefore it gets replaced. Whitespace is preserved after replacing.
+.if ${:Ua b b c:C,a b,,W} != " b c"
+.error
+.endif
+
+# The 'g' modifier does not have any effect here since each of the words
+# contains the character 'b' a single time.
+.if ${:Ua b b c:C,b,,g} != "a c"
+.error
+.endif
+
+# The first :C modifier has the 'W' modifier, which makes the whole
+# expression a single word. The 'g' modifier then replaces all occurrences
+# of "1 2" with "___". The 'W' modifier only applies to this single :C
+# modifier. This is demonstrated by the :C modifier that follows. If the
+# 'W' modifier would be preserved, only a single underscore would have been
+# replaced with an 'x'.
+.if ${:U1 2 3 1 2 3:C,1 2,___,Wg:C,_,x,} != "x__ 3 x__ 3"
+.error
+.endif
+
+# The regular expression does not match in the first word.
+# It matches once in the second word, and the \0\0 doubles that word.
+# In the third word, the regular expression matches as early as possible,
+# and since the matches must not overlap, the next possible match would
+# start at the 6, but at that point, there is only one character left,
+# and that cannot match the regular expression "..". Therefore only the
+# "45" is doubled in the result.
+.if ${:U1 23 456:C,..,\0\0,} != "1 2323 45456"
+.error
+.endif
+
+# The modifier '1' applies the replacement at most once, across the whole
+# variable value, no matter whether it is a single big word or many small
+# words.
+#
+# Up to 2020-08-28, the manual page said that the modifiers '1' and 'g'
+# were orthogonal, which was wrong.
+.if ${:U12345 12345:C,.,\0\0,1} != "112345 12345"
+.error
+.endif
+
+# Multiple asterisks form an invalid regular expression. This produces an
+# error message and (as of 2020-08-28) stops parsing in the middle of the
+# variable expression. The unparsed part of the expression is then copied
+# verbatim to the output, which is unexpected and can lead to strange shell
+# commands being run.
+mod-regex-compile-error:
+ @echo $@: ${:Uword1 word2:C,****,____,g:C,word,____,:Q}.
+
+# These tests generate error messages but as of 2020-08-28 just continue
+# parsing and execution as if nothing bad had happened.
mod-regex-limits:
- @echo $@:00-ok:${:U1 23 456:C,..,\0\0,:Q}
@echo $@:11-missing:${:U1 23 456:C,..,\1\1,:Q}
@echo $@:11-ok:${:U1 23 456:C,(.).,\1\1,:Q}
@echo $@:22-missing:${:U1 23 456:C,..,\2\2,:Q}
