diff --git a/Makefile.am b/Makefile.am
index 98a3adfd36..6b85f70251 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -17,6 +17,7 @@ bin_PROGRAMS = ctags
 if USE_READCMD
 bin_PROGRAMS+= readtags
 readtags_CPPFLAGS = -I. -I$(srcdir) -I$(srcdir)/main -I$(srcdir)/read
+readtags_CFLAGS = $(COVERAGE_CFLAGS)
 dist_readtags_SOURCES = $(READTAGS_SRCS) $(READTAGS_HEADS)
 readtags_CPPFLAGS += -DQUALIFIER -I$(srcdir)/dsl
 dist_readtags_SOURCES += $(QUALIFIER_SRCS) $(QUALIFIER_HEADS)
diff --git a/Units/roundtrip-escapes.d/expected.tags b/Units/roundtrip-escapes.d/expected.tags
new file mode 100644
index 0000000000..927ba14c21
--- /dev/null
+++ b/Units/roundtrip-escapes.d/expected.tags
@@ -0,0 +1,43 @@
+_\\_	input	//
+_\a_	input	//
+_\b_	input	//
+_\f_	input	//
+_\n_	input	//
+\n	input	//
+_\r_	input	//
+\r	input	//
+_\t_	input	//
+_\v_	input	//
+_\x01_	input	//
+_\x02_	input	//
+_\x03_	input	//
+_\x04_	input	//
+_\x05_	input	//
+_\x06_	input	//
+_\x07_	input	//
+_\x08_	input	//
+_\x09_	input	//
+_\x0A_	input	//
+_\x0b_	input	//
+_\x0c_	input	//
+_\x0d_	input	//
+_\x0e_	input	//
+_\x0f_	input	//
+_\x10_	input	//
+_\x11_	input	//
+_\x12_	input	//
+_\x13_	input	//
+_\x14_	input	//
+_\x15_	input	//
+_\x16_	input	//
+_\x17_	input	//
+_\x18_	input	//
+_\x19_	input	//
+_\x1a_	input	//
+_\x1b_	input	//
+_\x1c_	input	//
+_\x1d_	input	//
+_\x1e_	input	//
+_\x1F_	input	//
+_\x20_	input	//
+_\x21_	input	//
diff --git a/circle.yml b/circle.yml
index 45ce833134..ef2771a9f3 100644
--- a/circle.yml
+++ b/circle.yml
@@ -13,7 +13,7 @@ jobs:
        - run:
            name: Install build tools
            command: |
-             dnf -y install gcc automake autoconf pkgconfig bmake aspell-devel aspell-en libxml2-devel jansson-devel libyaml-devel || :
+             dnf -y install gcc automake autoconf pkgconfig bmake aspell-devel aspell-en libxml2-devel jansson-devel libyaml-devel findutils || :
        - run:
            name: Build
            command: |
diff --git a/makefiles/testing.mak b/makefiles/testing.mak
index f99e185076..5ceae60b89 100644
--- a/makefiles/testing.mak
+++ b/makefiles/testing.mak
@@ -1,7 +1,7 @@
 # -*- makefile -*-
 .PHONY: check units fuzz noise tmain tinst clean-units clean-tmain clean-gcov run-gcov codecheck cppcheck dicts cspell
 
-check: tmain units
+check: tmain units roundtrip
 
 clean-local: clean-units clean-tmain
 
@@ -18,13 +18,21 @@ LANGUAGES=
 CATEGORIES=
 UNITS=
 
+SILENT = $(SILENT_@AM_V@)
+SILENT_ = $(SILENT_@AM_DEFAULT_V@)
+SILENT_0 = @
+
+V_RUN = $(V_RUN_@AM_V@)
+V_RUN_ = $(V_RUN_@AM_DEFAULT_V@)
+V_RUN_0 = @echo "  RUN      $@";
+
 #
 # FUZZ Target
 #
 # SHELL must be dash or bash.
 #
 fuzz: $(CTAGS_TEST)
-	@ \
+	$(V_RUN) \
 	if test -n "$${ZSH_VERSION+set}"; then set -o SH_WORD_SPLIT; fi; \
 	if test x$(VG) = x1; then		\
 		VALGRIND=--with-valgrind;	\
@@ -40,7 +48,7 @@ fuzz: $(CTAGS_TEST)
 # NOISE Target
 #
 noise: $(CTAGS_TEST)
-	@ \
+	$(V_RUN) \
 	if test -n "$${ZSH_VERSION+set}"; then set -o SH_WORD_SPLIT; fi; \
 	if test x$(VG) = x1; then		\
 		VALGRIND=--with-valgrind;	\
@@ -56,7 +64,7 @@ noise: $(CTAGS_TEST)
 # CHOP Target
 #
 chop: $(CTAGS_TEST)
-	@ \
+	$(V_RUN) \
 	if test -n "$${ZSH_VERSION+set}"; then set -o SH_WORD_SPLIT; fi; \
 	if test x$(VG) = x1; then		\
 		VALGRIND=--with-valgrind;	\
@@ -68,7 +76,7 @@ chop: $(CTAGS_TEST)
 		--with-timeout=$(TIMEOUT)"; \
 	$(SHELL) $${c} $(srcdir)/Units
 slap: $(CTAGS_TEST)
-	@ \
+	$(V_RUN) \
 	if test -n "$${ZSH_VERSION+set}"; then set -o SH_WORD_SPLIT; fi; \
 	if test x$(VG) = x1; then		\
 		VALGRIND=--with-valgrind;	\
@@ -84,7 +92,7 @@ slap: $(CTAGS_TEST)
 # UNITS Target
 #
 units: $(CTAGS_TEST)
-	@ \
+	$(V_RUN) \
 	if test -n "$${ZSH_VERSION+set}"; then set -o SH_WORD_SPLIT; fi; \
 	if test x$(VG) = x1; then		\
 		VALGRIND=--with-valgrind;	\
@@ -115,8 +123,7 @@ clean-units:
 # Test main part, not parsers
 #
 tmain: $(CTAGS_TEST)
-	@ \
-	\
+	$(V_RUN) \
 	if test -n "$${ZSH_VERSION+set}"; then set -o SH_WORD_SPLIT; fi; \
 	if test x$(VG) = x1; then		\
 		VALGRIND=--with-valgrind;	\
@@ -144,22 +151,25 @@ clean-tmain:
 # Test installation
 #
 tinst:
-	@ \
-	\
+	$(V_RUN) \
 	builddir=$$(pwd); \
 	rm -rf $$builddir/$(TINST_ROOT); \
-	\
 	$(SHELL) $(srcdir)/misc/tinst $(srcdir) $$builddir/$(TINST_ROOT)
 
 #
 # Test readtags
 #
+if USE_READCMD
 roundtrip: $(READ_TEST)
-	@ \
-	\
+	$(V_RUN) \
 	builddir=$$(pwd); \
-	\
-	$(SHELL) $(srcdir)/misc/roundtrip $(READ_TEST) $${builddir}/Units
+	APPVEYOR=$(APPVEYOR) \
+		$(SHELL) $(srcdir)/misc/roundtrip $(READ_TEST) $${builddir}/Units
+else
+# apparently Automake doesn't like adding an extra dependency in a conditional,
+# so add it unconditionally and simply have it not do anything if it's disabled.
+roundtrip:
+endif
 
 #
 # Checking code in ctags own rules
diff --git a/misc/roundtrip b/misc/roundtrip
index ae356aeda3..8ce5685caf 100644
--- a/misc/roundtrip
+++ b/misc/roundtrip
@@ -15,15 +15,15 @@
 #   You should have received a copy of the GNU General Public License
 #   along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
+set -e
+
+# Avoid trouble with weird bytes on non-C locales
+export LC_ALL=C
+
 READTAGS=${1:-./readtags}
 UNITS=${2:-./Units}
 s=0
 
-sink()
-{
-    cat > /dev/null
-}
-
 if ! [ -f "${READTAGS}" ]; then
     echo "No such file: ${READTAGS}" 1>&2
     exit 1
@@ -40,16 +40,116 @@ if ! [ -x "${UNITS}" ]; then
 fi
 
 
-for tags in $(find "$UNITS" -name expected.tags); do
-    for t in $(sed -e 's/^\([^\t]*\)\t.*/\1/' "$tags"); do
-	if r=$( "${READTAGS}" -t "$tags" - "$t" 2>/dev/null ); then
-	    :
-	else
-	    printf "FAILED[%3d]: " $?
-	    echo "${READTAGS}" -t "$tags" - "$t"
+# Expand CTags escape sequences (and keep the original, too)
+# See docs/format.rst
+expandEscapeSequences()
+{
+    sed -e '
+# early out if there is no escape sequence (common case)
+/\\/!b
+
+# loop until we handled all occurrences of \.
+# as it is not possible to do all replacements at once and we need not to
+# ever use the result of an replacement to perform another one, use two
+# reserved placeholders.
+:again
+    s/\\/__BACKSLASH__/
+    s/__BACKSLASH__\\/__LITBACKSLASH__/;	t again
+    s/__BACKSLASH__t/	/;			t again
+    s/__BACKSLASH__r/'"$(printf '\r')"'/;	t again
+    s/__BACKSLASH__n/\
+/;						t again
+    s/__BACKSLASH__a//;			t again
+    s/__BACKSLASH__b/'"$(printf '\b')"'/;	t again
+    s/__BACKSLASH__v//;			t again
+    s/__BACKSLASH__f//;			t again
+    s/__BACKSLASH__x01/
/;			t again
+    s/__BACKSLASH__x02//;			t again
+    s/__BACKSLASH__x03//;			t again
+    s/__BACKSLASH__x04//;			t again
+    s/__BACKSLASH__x05//;			t again
+    s/__BACKSLASH__x06//;			t again
+    s/__BACKSLASH__x07//;			t again
+    s/__BACKSLASH__x08//;			t again
+    s/__BACKSLASH__x09/	/;			t again
+    s/__BACKSLASH__x0[aA]/\
+/;						t again
+    s/__BACKSLASH__x0[bB]//;			t again
+    s/__BACKSLASH__x0[cC]//;			t again
+    s/__BACKSLASH__x0[dD]/'"$(printf '\r')"'/;	t again
+    s/__BACKSLASH__x0[eE]//;			t again
+    s/__BACKSLASH__x0[fF]//;			t again
+    s/__BACKSLASH__x10//;			t again
+    s/__BACKSLASH__x11//;			t again
+    s/__BACKSLASH__x12//;			t again
+    s/__BACKSLASH__x13//;			t again
+    s/__BACKSLASH__x14//;			t again
+    s/__BACKSLASH__x15//;			t again
+    s/__BACKSLASH__x16//;			t again
+    s/__BACKSLASH__x17//;			t again
+    s/__BACKSLASH__x18//;			t again
+    s/__BACKSLASH__x19//;			t again
+    s/__BACKSLASH__x1[aA]//;			t again
+    s/__BACKSLASH__x1[bB]//;			t again
+    s/__BACKSLASH__x1[cC]//;			t again
+    s/__BACKSLASH__x1[dD]//;			t again
+    s/__BACKSLASH__x1[eE]//;			t again
+    s/__BACKSLASH__x1[fF]//;			t again
+    s/__BACKSLASH__x21/!/;			t again
+    s/__BACKSLASH__x20/ /;			t again
+    s/__BACKSLASH__x7[fF]//;			t again
+    /\\/b again
+
+:out
+    # replace lonely "\"es
+    s/__BACKSLASH__/\\/g
+    # replace literal "\"es ("\\" from the input)
+    s/__LITBACKSLASH__/\\/g
+'
+}
+
+# Escape characters MSYS2 would convert when calling readtags
+if [ -n "$APPVEYOR" ]; then
+    MSYS2_READTAGS="${READTAGS}"
+    msys2_readtags()
+    {
+	MSYS2_ARG_CONV_EXCL="$4" \
+	"${MSYS2_READTAGS}" "$@"
+    }
+    READTAGS=msys2_readtags
+fi
+
+# disable path expansion, because we don't need it and it's applied on the
+# result of the commands used to loop on, and we don't want that.
+set -f
+# remove space from IFS as it's valid in tag names
+OLD_IFS="$IFS"
+IFS='
+'
+
+tagfiles="$(find "$UNITS" -name expected.tags)"
+for tags in $tagfiles; do
+    tagnames="$(sed -e 's/^\([^	]*\)	.*/\1/' "$tags")"
+    for name in $tagnames; do
+	# Yes, there is a reason for this craziness.  We need to properly
+	# handle embedded newlines (expanded from "\n"), including trailing
+	# ones the shell would strip automatically.  To work around this, we
+	# add a dummy character at the end to inhibit stripping, and then
+	# remove it, plus the extra newline, using variable substitutions.
+	# Note: we use "printf '%s\n'" instead of "echo" because Dash's "echo"
+	# unconditionally expands some sequences, like "\t" and alike.
+	t="$(printf '%s\n' "$name" | expandEscapeSequences; printf _)"
+	t="${t%
+_}"
+
+	if [ 1 -gt $("${READTAGS}" -t "$tags" - "$t" | wc -l) ]; then
+	    printf 'FAILED: "%s" -t "%s" - "%s"\n' "${READTAGS}" "$tags" "$t"
+	    printf '	The raw tag name was "%s"\n' "$name"
 	    s=1
 	fi
-    done 
+    done
 done
 
+IFS="$OLD_IFS"
+
 exit $s
diff --git a/read/readtags-cmd.c b/read/readtags-cmd.c
index 4722995557..873237df56 100644
--- a/read/readtags-cmd.c
+++ b/read/readtags-cmd.c
@@ -95,6 +95,11 @@ static void findTag (const char *const name, const int options)
 				printTag (&entry);
 			} while (tagsFindNext (file, &entry) == TagSuccess);
 		}
+		else
+		{
+			fprintf (stderr, "%s: no match found for \"%s\" in \"%s\"\n",
+					ProgramName, name, TagFileName);
+		}
 		tagsClose (file);
 	}
 }
diff --git a/read/readtags.c b/read/readtags.c
index 7031087ecc..5ec0483a10 100644
--- a/read/readtags.c
+++ b/read/readtags.c
@@ -93,6 +93,59 @@ static const char *const PseudoTagPrefix = "!_";
 *   FUNCTION DEFINITIONS
 */
 
+/* Converts a hexadecimal digit to its value */
+static int xdigitValue (char digit)
+{
+	if (digit >= '0' && digit <= '9')
+		return digit - '0';
+	else if (digit >= 'a' && digit <= 'f')
+		return 10 + digit - 'a';
+	else if (digit >= 'A' && digit <= 'F')
+		return 10 + digit - 'A';
+	else
+		return 0;
+}
+
+/*
+ * Reads the first character from the string, possibly un-escaping it, and
+ * advances *s to the start of the next character.
+ */
+static int readTagCharacter (const char **s)
+{
+	int c = **s;
+
+	(*s)++;
+
+	if (c == '\\')
+	{
+		switch (**s)
+		{
+			case 't': c = '\t'; (*s)++; break;
+			case 'r': c = '\r'; (*s)++; break;
+			case 'n': c = '\n'; (*s)++; break;
+			case '\\': c = '\\'; (*s)++; break;
+			/* Universal-CTags extensions */
+			case 'a': c = '\a'; (*s)++; break;
+			case 'b': c = '\b'; (*s)++; break;
+			case 'v': c = '\v'; (*s)++; break;
+			case 'f': c = '\f'; (*s)++; break;
+			case 'x':
+				if (isxdigit ((*s)[1]) && isxdigit ((*s)[2]))
+				{
+					int val = (xdigitValue ((*s)[1]) << 4) | xdigitValue ((*s)[2]);
+					if (val < 0x80)
+					{
+						(*s) += 3;
+						c = val;
+					}
+				}
+				break;
+		}
+	}
+
+	return c;
+}
+
 /*
  * Compare two strings, ignoring case.
  * Return 0 for match, < 0 for smaller, > 0 for bigger
@@ -100,23 +153,59 @@ static const char *const PseudoTagPrefix = "!_";
  * This makes a difference when one of the chars lies between upper and lower
  * ie. one of the chars [ \ ] ^ _ ` for ascii. (The '_' in particular !)
  */
-static int struppercmp (const char *s1, const char *s2)
+static int taguppercmp (const char *s1, const char *s2)
+{
+	int result;
+	int c1, c2;
+	do
+	{
+		c1 = *s1++;
+		c2 = readTagCharacter (&s2);
+
+		result = toupper (c1) - toupper (c2);
+	} while (result == 0  &&  c1 != '\0'  &&  c2 != '\0');
+	return result;
+}
+
+static int tagnuppercmp (const char *s1, const char *s2, size_t n)
+{
+	int result;
+	int c1, c2;
+	do
+	{
+		c1 = *s1++;
+		c2 = readTagCharacter (&s2);
+
+		result = toupper (c1) - toupper (c2);
+	} while (result == 0  &&  --n > 0  &&  c1 != '\0'  &&  c2 != '\0');
+	return result;
+}
+
+static int tagcmp (const char *s1, const char *s2)
 {
 	int result;
+	int c1, c2;
 	do
 	{
-		result = toupper ((int) *s1) - toupper ((int) *s2);
-	} while (result == 0  &&  *s1++ != '\0'  &&  *s2++ != '\0');
+		c1 = *s1++;
+		c2 = readTagCharacter (&s2);
+
+		result = c1 - c2;
+	} while (result == 0  &&  c1 != '\0'  &&  c2 != '\0');
 	return result;
 }
 
-static int strnuppercmp (const char *s1, const char *s2, size_t n)
+static int tagncmp (const char *s1, const char *s2, size_t n)
 {
 	int result;
+	int c1, c2;
 	do
 	{
-		result = toupper ((int) *s1) - toupper ((int) *s2);
-	} while (result == 0  &&  --n > 0  &&  *s1++ != '\0'  &&  *s2++ != '\0');
+		c1 = *s1++;
+		c2 = readTagCharacter (&s2);
+
+		result = c1 - c2;
+	} while (result == 0  &&  --n > 0  &&  c1 != '\0'  &&  c2 != '\0');
 	return result;
 }
 
@@ -311,6 +400,7 @@ static void parseTagLine (tagFile *file, tagEntry *const entry)
 {
 	int i;
 	char *p = file->line.buffer;
+	size_t p_len = strlen (p);
 	char *tab = strchr (p, TAB);
 
 	entry->fields.list = NULL;
@@ -320,8 +410,25 @@ static void parseTagLine (tagFile *file, tagEntry *const entry)
 
 	entry->name = p;
 	if (tab != NULL)
-	{
 		*tab = '\0';
+	while (*p != '\0')
+	{
+		const char *next = p;
+		int ch = readTagCharacter (&next);
+		size_t skip = next - p;
+
+		*p = (char) ch;
+		p++;
+		p_len -= skip;
+		if (skip > 1)
+		{
+			memmove (p, next, p_len);
+			tab -= skip - 1;
+		}
+	}
+
+	if (tab != NULL)
+	{
 		p = tab + 1;
 		entry->file = p;
 		tab = strchr (p, TAB);
@@ -561,18 +668,18 @@ static int nameComparison (tagFile *const file)
 	if (file->search.ignorecase)
 	{
 		if (file->search.partial)
-			result = strnuppercmp (file->search.name, file->name.buffer,
+			result = tagnuppercmp (file->search.name, file->name.buffer,
 					file->search.nameLength);
 		else
-			result = struppercmp (file->search.name, file->name.buffer);
+			result = taguppercmp (file->search.name, file->name.buffer);
 	}
 	else
 	{
 		if (file->search.partial)
-			result = strncmp (file->search.name, file->name.buffer,
+			result = tagncmp (file->search.name, file->name.buffer,
 					file->search.nameLength);
 		else
-			result = strcmp (file->search.name, file->name.buffer);
+			result = tagcmp (file->search.name, file->name.buffer);
 	}
 	return result;
 }