awk: fix subst code to handle "start of word" pattern correctly (needs REG_STARTEND)

function                                             old     new   delta
awk_sub                                              637     714     +77

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

1 files changed, 15 insertions, 13 deletions

diff --git a/testsuite/awk.tests b/testsuite/awk.tests
index c61d32947..5a792c241 100755
--- a/testsuite/awk.tests
+++ b/testsuite/awk.tests
@@ -557,14 +557,12 @@ testing 'awk gensub backslashes \' \
 	'awk '$sq'BEGIN { s="\\"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\
 \\|\\
-' \
-	'' ''
+' '' ''
 testing 'awk gensub backslashes \\' \
 	'awk '$sq'BEGIN { s="\\\\"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\\\
 \\|\\
-' \
-	'' ''
+' '' ''
 # gawk 5.1.1 handles trailing unpaired \ inconsistently.
 # If replace string is single \, it is used verbatim,
 # but if it is \\\ (three slashes), gawk uses "\<NUL>" (!!!), not "\\" as you would expect.
@@ -572,31 +570,35 @@ testing 'awk gensub backslashes \\\' \
 	'awk '$sq'BEGIN { s="\\\\\\"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\\\\\
 \\\\|\\\\
-' \
-	'' ''
+' '' ''
 testing 'awk gensub backslashes \\\\' \
 	'awk '$sq'BEGIN { s="\\\\\\\\"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\\\\\\\
 \\\\|\\\\
-' \
-	'' ''
+' '' ''
 testing 'awk gensub backslashes \&' \
 	'awk '$sq'BEGIN { s="\\&"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\&
 &|&
-' \
-	'' ''
+' '' ''
 testing 'awk gensub backslashes \0' \
 	'awk '$sq'BEGIN { s="\\0"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\0
 a|a
-' \
-	'' ''
+' '' ''
 testing 'awk gensub backslashes \\0' \
 	'awk '$sq'BEGIN { s="\\\\0"; print "s=" s; print gensub("a", s, "g", "a|a") }'$sq \
 	's=\\\\0
 \\0|\\0
-' \
+' '' ''
+
+# The "b" in "abc" should not match <b* pattern.
+# Currently we use REG_STARTEND ("This flag is a BSD extension, not present in POSIX")
+# to implement the code to handle this correctly, but if your libc has no REG_STARTEND,
+# the alternative code mishandles this case.
+testing 'awk gsub erroneous word start match' \
+	"awk 'BEGIN { a=\"abc\"; gsub(/\<b*/,\"\",a); print a }'" \
+	'abc\n' \
 	'' ''
 
 exit $FAILCOUNT