grep: option to use GNU regex matching instead of POSIX one.
This fixes problems with NULs in files being scanned, but
costs +800 bytes. The same can be done to sed (TODO).
diff --git a/testsuite/grep.tests b/testsuite/grep.tests
index b2de2af..8cee1b9 100755
--- a/testsuite/grep.tests
+++ b/testsuite/grep.tests
@@ -62,12 +62,8 @@
"grep -s domatch nonexistent - ; echo \$?" \
"(standard input):domatch\n2\n" "" "nomatch\ndomatch\nend\n"
-# This doesn't match GNU behaviour (Binary file input matches)
-# acts like GNU grep -a
-testing "grep handles binary files" "grep foo input" "foo\n" "\0foo\n\n" ""
-# This doesn't match GNU behaviour (Binary file (standard input) matches)
-# acts like GNU grep -a
-testing "grep handles binary stdin" "grep foo" "foo\n" "" "\0foo\n\n"
+testing "grep handles NUL in files" "grep -a foo input" "\0foo\n" "\0foo\n\n" ""
+testing "grep handles NUL on stdin" "grep -a foo" "\0foo\n" "" "\0foo\n\n"
testing "grep matches NUL" "grep . input > /dev/null 2>&1 ; echo \$?" \
"0\n" "\0\n" ""