First commit

Signed-off-by: Mattias Andrée <maandree@kth.se>

5 files changed, 734 insertions, 0 deletions

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..0ec616b
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,15 @@
+*\#*
+*~
+*.o
+*.a
+*.lo
+*.su
+*.so
+*.so.*
+*.dll
+*.dylib
+*.gch
+*.gcov
+*.gcno
+*.gcda
+/git-rediff
diff --git a/LICENSE b/LICENSE
new file mode 100644
index 0000000..fccd785
--- /dev/null
+++ b/LICENSE
@@ -0,0 +1,15 @@
+ISC License
+
+© 2024 Mattias Andrée <maandree@kth.se>
+
+Permission to use, copy, modify, and/or distribute this software for any
+purpose with or without fee is hereby granted, provided that the above
+copyright notice and this permission notice appear in all copies.
+
+THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
diff --git a/Makefile b/Makefile
new file mode 100644
index 0000000..803a6e7
--- /dev/null
+++ b/Makefile
@@ -0,0 +1,37 @@
+.POSIX:
+
+CONFIGFILE = config.mk
+include $(CONFIGFILE)
+
+OBJ =\
+	git-rediff.o
+
+HDR =
+
+all: git-rediff
+$(OBJ): $(HDR)
+
+.c.o:
+	$(CC) -c -o $@ $< $(CFLAGS) $(CPPFLAGS)
+
+git-rediff: $(OBJ)
+	$(CC) -o $@ $(OBJ) $(LDFLAGS)
+
+install: git-rediff
+	mkdir -p -- "$(DESTDIR)$(PREFIX)/bin"
+	mkdir -p -- "$(DESTDIR)$(MANPREFIX)/man1/"
+	cp -- git-rediff "$(DESTDIR)$(PREFIX)/bin/"
+	cp -- git-rediff.1 "$(DESTDIR)$(MANPREFIX)/man1/"
+
+uninstall:
+	-rm -f -- "$(DESTDIR)$(PREFIX)/bin/git-rediff"
+	-rm -f -- "$(DESTDIR)$(MANPREFIX)/man1/git-rediff.1"
+
+clean:
+	-rm -f -- *.o *.a *.lo *.su *.so *.so.* *.gch *.gcov *.gcno *.gcda
+	-rm -f -- git-rediff
+
+.SUFFIXES:
+.SUFFIXES: .o .c
+
+.PHONY: all install uninstall clean
diff --git a/config.mk b/config.mk
new file mode 100644
index 0000000..aeb8cec
--- /dev/null
+++ b/config.mk
@@ -0,0 +1,8 @@
+PREFIX    = /usr
+MANPREFIX = $(PREFIX)/share/man
+
+CC = c99
+
+CPPFLAGS = -D_DEFAULT_SOURCE -D_BSD_SOURCE -D_XOPEN_SOURCE=700 -D_GNU_SOURCE
+CFLAGS   =
+LDFLAGS  = -lsimple
diff --git a/git-rediff.c b/git-rediff.c
new file mode 100644
index 0000000..07629ef
--- /dev/null
+++ b/git-rediff.c
@@ -0,0 +1,659 @@
+/* See LICENSE file for copyright and license details. */
+#ifdef __linux__
+# include <linux/prctl.h>
+# include <sys/prctl.h>
+#endif
+#include <libsimple.h>
+#include <libsimple-arg.h>
+
+USAGE("[file] ...");
+
+#if defined(__clang__)
+# pragma clang diagnostic ignored "-Wunsafe-buffer-usage"
+#endif
+
+enum successfulness {
+	MERGED = 0,
+	CONFLICT = 1,
+	ERROR = 2
+};
+
+struct line {
+	union {
+		size_t in_off;
+		const unsigned char *in;
+	};
+	const char *text;
+	size_t len;
+};
+
+struct text {
+	struct line *lines;
+	size_t nlines;
+	size_t lines_size;
+};
+
+struct subhunk {
+	union {
+		struct line *head;
+		unsigned char *in;
+	};
+	struct text text;
+};
+
+struct hunk {
+	struct subhunk *subs;
+	size_t nsubs;
+};
+
+static int
+send_line(int fd, const char *fname, const struct line *line)
+{
+	size_t off = 0;
+	ssize_t r;
+	while (off < line->len) {
+		r = write(fd, &line->text[off], line->len - off);
+		if (r < 0) {
+			if (errno == EINTR)
+				continue;
+			weprintf("write %s:", fname);
+			return -1;
+		}
+		off += (size_t)r;
+	}
+	return 0;
+}
+
+static int
+send_text(int fd, const char *fname, const struct text *text)
+{
+	size_t i;
+	for (i = 0; i < text->nlines; i++)
+		if (send_line(fd, fname, &text->lines[i]))
+			return -1;
+	return 0;
+}
+
+static void
+ensure_nonstandard(int *fdp)
+{
+	int r;
+	if (*fdp > 2)
+		return;
+	r = fcntl(*fdp, F_DUPFD, 3);
+	if (r < 3)
+		eprintf("fcntl <pipe> F_DUPFD 3:");
+	*fdp = r;
+}
+
+#if defined(__GNUC__)
+__attribute__((__pure__))
+#endif
+static int
+in_all(const unsigned char *in, size_t full_bytes, unsigned char last_byte)
+{
+	while (full_bytes--)
+		if (!(~*in++ & ((1U << CHAR_BIT) - 1U)))
+			return 0;
+	return *in == last_byte;
+}
+
+static int
+line_startswith(struct line *line, const char *head)
+{
+	size_t len = strlen(head);
+	return line->len >= len && !strncmp(line->text, head, len);
+}
+
+static void
+append_lines(struct text *textp, const struct line *lines, size_t n)
+{
+	if (n > textp->lines_size - textp->nlines) {
+		textp->lines_size = textp->nlines + n;
+		textp->lines = ereallocarray(textp->lines, textp->lines_size, sizeof(*textp->lines));
+	}
+	memcpy(&textp->lines[textp->nlines], lines, n * sizeof(*lines));
+	textp->nlines += n;
+}
+
+static void
+append_line(struct text *textp, const struct line *line)
+{
+	append_lines(textp, line, 1U);
+}
+
+static void
+append_text(struct text *textp, const struct text *text)
+{
+	append_lines(textp, text->lines, text->nlines);
+}
+
+static char *
+diff_subhunks(const struct subhunk *f1, const struct subhunk *f2)
+{
+	size_t context = MAX(f1->text.nlines, f2->text.nlines);
+	pid_t pid1, pid2, piddiff;
+	int pipe1[2], pipe2[2], pipediff[2];
+	char *ret = NULL;
+	size_t ret_size = 0;
+	size_t ret_off = 0;
+	int status;
+	size_t i, n;
+	int rem, ret_next = 0;
+	ssize_t r;
+	union {
+		struct {
+			char context[sizeof("-U") + 3 * sizeof(context)];
+			char pipe1[sizeof("/dev/fd/-") + 3 * sizeof(pipe1[0])];
+			char pipe2[sizeof("/dev/fd/-") + 3 * sizeof(pipe2[0])];
+		};
+		char text[8096];
+	} buf;
+
+	if (pipe(pipediff))
+		eprintf("pipe:");
+	if (pipe(pipe1))
+		eprintf("pipe:");
+	if (pipe(pipe2))
+		eprintf("pipe:");
+
+	ensure_nonstandard(&pipe2[0]);
+	ensure_nonstandard(&pipe1[0]);
+
+	pid1 = fork();
+	if (pid1 < 0)
+		eprintf("fork:");
+	if (pid1 == 0) {
+		close(pipediff[0]);
+		close(pipediff[1]);
+		close(pipe1[0]);
+		close(pipe2[0]);
+		close(pipe2[1]);
+		if (send_text(pipe1[1], "<pipe>", &f1->text))
+			_exit(1);
+		close(pipe1[1]);
+		_exit(0);
+	}
+
+	pid2 = fork();
+	if (pid2 < 0)
+		eprintf("fork:");
+	if (pid2 == 0) {
+		close(pipediff[0]);
+		close(pipediff[1]);
+		close(pipe2[0]);
+		close(pipe1[0]);
+		close(pipe1[1]);
+		if (send_text(pipe2[1], "<pipe>", &f2->text))
+			_exit(1);
+		close(pipe2[1]);
+		_exit(0);
+	}
+
+	close(pipe1[1]);
+	close(pipe2[1]);
+
+	piddiff = fork();
+	if (piddiff < 0)
+		eprintf("fork:");
+	if (piddiff == 0) {
+		close(pipediff[0]);
+		if (pipediff[1] != STDOUT_FILENO) {
+			if (dup2(pipediff[1], STDOUT_FILENO) != STDOUT_FILENO)
+				eprintf("dup2 <pipe> <stdout>:");
+			close(pipediff[1]);
+		}
+		sprintf(buf.context, "-U%zu", context);
+		sprintf(buf.pipe1, "/dev/fd/%i", pipe1[0]);
+		sprintf(buf.pipe2, "/dev/fd/%i", pipe2[0]);
+#ifdef PR_SET_PDEATHSIG
+		prctl(PR_SET_PDEATHSIG, SIGKILL);
+#endif
+		execlp("diff", "diff", buf.context, "--", buf.pipe1, buf.pipe2, NULL);
+		_exit(125);
+	}
+
+	close(pipediff[1]);
+	close(pipe1[0]);
+	close(pipe2[0]);
+
+	rem = 3;
+	for (;;) {
+		r = read(pipediff[0], buf.text, sizeof(buf.text));
+		if (r <= 0) {
+			if (!r)
+				break;
+			if (errno == EINTR)
+				continue;
+			eprintf("read <diff(1) subprocess stdout pipe>:");
+		}
+		n = (size_t)r;
+		for (i = 0; i < n; i++) {
+			if (ret_next) {
+				ret_next = 0;
+				if (ret_off == ret_size) {
+					if (ret_size > SIZE_MAX - 512) {
+						errno = ENOMEM;
+						eprintf("realloc:");
+					}
+					ret = erealloc(ret, ret_size += 512);
+				}
+				ret[ret_off++] = buf.text[i];
+			} else if (buf.text[i] == '\n') {
+				if (rem) {
+					if (!--rem)
+						ret_next = 1;
+					continue;
+				}
+				ret_next = 1;
+			}
+		}
+	}
+
+	if (waitpid(pid1, &status, 0) != pid1)
+		eprintf("waitpid <file sender subprocess> 0:");
+	if (status)
+		eprintf("waitpid <file sender subprocess> 0: process exited abnormally");
+
+	if (waitpid(pid2, &status, 0) != pid2)
+		eprintf("waitpid <file sender subprocess> 0:");
+	if (status)
+		eprintf("waitpid <file sender subprocess> 0: process exited abnormally");
+
+	if (waitpid(piddiff, &status, 0) != piddiff)
+		eprintf("waitpid <diff(1) subprocess> 0:");
+	if (status == 0) {
+		ret = erealloc(ret, f1->text.nlines + 1U);
+		memset(ret, ' ', f1->text.nlines);
+	} else if (WIFEXITED(status) && WEXITSTATUS(status) == 1) {
+		ret = erealloc(ret, ret_off + 1U);
+	} else {
+		eprintf("waitpid <diff(1) subprocess> 0: process exited abnormally");
+	}
+	ret[ret_off] = '\0';
+	return ret;
+}
+
+static void
+diff_hunk(struct subhunk *ret, const struct hunk *hunk)
+{
+	char *diff, *p;
+	size_t k, i, j, in_size = 0, in_off = 0;
+	size_t in_step, bit_off;
+	unsigned char bit;
+	struct subhunk ret_buf, ret_tmp;
+
+	in_step = hunk->nsubs / CHAR_BIT;
+	in_step += (size_t)!!(hunk->nsubs % CHAR_BIT);
+
+	bit = (unsigned char)1;
+	bit_off = 0;
+	ret->text.lines_size = ret->text.nlines = hunk->subs[0].text.nlines;
+	in_size = ret->text.nlines * in_step;
+	ret->in = emalloc(in_size);
+	ret->text.lines = ecalloc(ret->text.lines_size, sizeof(*ret->text.lines));
+	memcpy(ret->text.lines, hunk->subs[0].text.lines, ret->text.nlines * sizeof(*hunk->subs[0].text.lines));
+	memset(ret->in, 0, in_size);
+	for (i = 0; i < ret->text.nlines; i++) {
+		ret->text.lines[i].in_off = in_off;
+		ret->in[in_off + bit_off] = bit;
+		in_off += in_step;
+	}
+
+	ret_buf.in = ret->in;
+	ret_buf.text.lines = NULL;
+	ret_buf.text.lines_size = 0;
+
+	for (k = 1; k < hunk->nsubs; k++) {
+		bit <<= 1;
+		if (!bit) {
+			bit = (unsigned char)1;
+			bit_off += 1;
+		}
+		diff = diff_subhunks(ret, &hunk->subs[k]);
+		i = j = 0;
+		ret_buf.text.nlines = 0;
+		for (p = diff; *p; p++) {
+			if (*p == '-') {
+				append_line(&ret_buf.text, &ret->text.lines[i]);
+				i++;
+			} else if (*p == '+') {
+				append_line(&ret_buf.text, &hunk->subs[k].text.lines[j]);
+				if (in_off == in_size) {
+					if (in_step > (SIZE_MAX - in_size) / 16U) {
+						errno = ENOMEM;
+						eprintf("realloc:");
+					}
+					in_size += 16U * in_step;
+					ret_buf.in = ret->in = erealloc(ret->in, in_size);
+					memset(&ret->in[in_off], 0, in_size - in_off);
+				}
+				ret_buf.text.lines[ret_buf.text.nlines - 1U].in_off = in_off;
+				ret_buf.in[ret_buf.text.lines[ret_buf.text.nlines - 1U].in_off + bit_off] = bit;
+				in_off += in_step;
+				j++;
+			} else if (*p == ' ') {
+				append_line(&ret_buf.text, &ret->text.lines[i]);
+				ret_buf.in[ret_buf.text.lines[ret_buf.text.nlines - 1U].in_off + bit_off] |= bit;
+				i++;
+				j++;
+			} else {
+				eprintf("output of diff(1) was corrupted");
+			}
+		}
+		ret->text.nlines = ret_buf.text.nlines;
+		ret_tmp.text = ret->text;
+		ret->text = ret_buf.text;
+		ret_buf.text = ret_tmp.text;
+		free(diff);
+	}
+
+	for (i = 0; i < ret->text.nlines; i++)
+		ret->text.lines[i].in = &ret->in[ret->text.lines[i].in_off];
+
+	free(ret_buf.text.lines);
+}
+
+static enum successfulness
+rediff_hunk(struct text *resp, const struct hunk *hunk, const struct line *tail)
+{
+	struct subhunk diff;
+	size_t i, j, full_bytes;
+	unsigned char last_byte;
+	struct hunk uncommon;
+	int in_uncommon = 0;
+	enum successfulness ret = MERGED;
+
+	full_bytes = hunk->nsubs / CHAR_BIT;
+	last_byte = 0;
+	if (hunk->nsubs % CHAR_BIT) {
+		last_byte = (unsigned char)(1U << ((hunk->nsubs % CHAR_BIT) - 1U));
+		last_byte |= (unsigned char)(last_byte - 1U);
+	}
+
+	uncommon.nsubs = hunk->nsubs;
+	uncommon.subs = ecalloc(uncommon.nsubs, sizeof(*uncommon.subs));
+	for (i = 0; i < uncommon.nsubs; i++) {
+		uncommon.subs[i].head = hunk->subs[i].head;
+		uncommon.subs[i].text.lines = NULL;
+		uncommon.subs[i].text.nlines = 0;
+		uncommon.subs[i].text.lines_size = 0;
+	}
+
+	diff_hunk(&diff, hunk);
+	for (i = 0; i < diff.text.nlines; i++) {
+		if (in_all(diff.text.lines[i].in, full_bytes, last_byte)) {
+			if (in_uncommon) {
+				for (j = 0; j < uncommon.nsubs; j++) {
+					append_line(resp, uncommon.subs[j].head);
+					append_text(resp, &uncommon.subs[j].text);
+				}
+				append_line(resp, tail);
+				in_uncommon = 0;
+			}
+			append_line(resp, &diff.text.lines[i]);
+		} else {
+			if (!in_uncommon) {
+				in_uncommon = 1;
+				for (j = 0; j < uncommon.nsubs; j++)
+					uncommon.subs[j].text.nlines = 0;
+			}
+			ret = CONFLICT;
+			for (j = 0; j < uncommon.nsubs; j++)
+				if ((diff.text.lines[i].in[j / CHAR_BIT] >> (j % CHAR_BIT)) & 1U)
+					append_line(&uncommon.subs[j].text, &diff.text.lines[i]);
+		}
+	}
+
+	if (in_uncommon) {
+		for (j = 0; j < uncommon.nsubs; j++) {
+			append_line(resp, uncommon.subs[j].head);
+			append_text(resp, &uncommon.subs[j].text);
+		}
+		append_line(resp, tail);
+	}
+
+	free(diff.in);
+	free(diff.text.lines);
+	for (i = 0; i < uncommon.nsubs; i++)
+		free(uncommon.subs[i].text.lines);
+	free(uncommon.subs);
+
+	return ret;
+}
+
+static enum successfulness
+rediff_file(struct text *text_out, const struct text *text_in, const char *fname)
+{
+	size_t i, t;
+	struct hunk hunk = {0};
+	ssize_t subhunk = -1;
+	enum successfulness ret = MERGED, r;
+
+	*text_out = (struct text){0};
+
+	for (i = 0; i < text_in->nlines; i++) {
+		if (line_startswith(&text_in->lines[i], "<<<<<<<")) {
+			if (subhunk >= 0)
+				goto syntax_error;
+			goto new_subhunk;
+		} else if (line_startswith(&text_in->lines[i], "|||||||") || line_startswith(&text_in->lines[i], "=======")) {
+			if (subhunk < 0)
+				goto syntax_error;
+			if (!line_startswith(hunk.subs[subhunk].head, "<<<<<<<") &&
+			    !line_startswith(hunk.subs[subhunk].head, "|||||||"))
+				goto syntax_error;
+		new_subhunk:
+			subhunk++;
+			if ((size_t)subhunk == hunk.nsubs)
+				hunk.subs = ereallocarray(hunk.subs, ++hunk.nsubs, sizeof(*hunk.subs));
+			hunk.subs[subhunk].text.nlines = 0;
+			hunk.subs[subhunk].head = &text_in->lines[i];
+		} else if (line_startswith(&text_in->lines[i], ">>>>>>>")) {
+			if (subhunk < 0)
+				goto syntax_error;
+			if (!line_startswith(hunk.subs[subhunk].head, "======="))
+				goto syntax_error;
+			t = hunk.nsubs;
+			hunk.nsubs = (size_t)subhunk + 1U;
+			r = rediff_hunk(text_out, &hunk, &text_in->lines[i]);
+			ret = MAX(ret, r);
+			hunk.nsubs = t;
+			subhunk = -1;
+		} else {
+			if (subhunk < 0)
+				append_line(text_out, &text_in->lines[i]);
+			else
+				append_line(&hunk.subs[subhunk].text, &text_in->lines[i]);
+		}
+	}
+
+	if (subhunk >= 0) {
+		weprintf("file %s is truncated", fname);
+	error:
+		ret = ERROR;
+	}
+
+	for (i = 0; i < hunk.nsubs; i++)
+		free(hunk.subs[i].text.lines);
+	free(hunk.subs);
+
+	return ret;
+
+syntax_error:
+	weprintf("syntax error at %s:%zu", fname, i + 1U);
+	goto error;
+}
+
+static int
+read_lines(struct text *lines_out, char **text_out, int fd, const char *fname)
+{
+	struct line *lines;
+	size_t text_len = 0;
+	size_t text_size = 0;
+	ssize_t r;
+	void *new;
+	size_t i;
+
+	*lines_out = (struct text){0};
+	*text_out = NULL;
+
+	for (;;) {
+		if (text_len == text_size) {
+			if (text_size > SIZE_MAX - 8096U) {
+				errno = ENOMEM;
+				weprintf("realloc:");
+				goto fail;
+			}
+			text_size += 8096U;
+			new = realloc(*text_out, text_size);
+			if (!new)
+				goto fail;
+			*text_out = new;
+		}
+		r = read(fd, &(*text_out)[text_len], text_size - text_len);
+		if (r <= 0) {
+			if (!r)
+				break;
+			if (errno == EINTR)
+				continue;
+			weprintf("read %s:", fname);
+			goto fail;
+		}
+		text_len += (size_t)r;
+	}
+
+	lines_out->nlines = (text_len ? 1U : 0U);
+	for (i = 0; i + 1U < text_len; i++)
+		if ((*text_out)[i] == '\n')
+			lines_out->nlines += 1U;
+
+	if (!lines_out->nlines)
+		return 0;
+
+	lines = lines_out->lines = ecalloc(lines_out->nlines, sizeof(*lines_out->lines));
+
+	for (i = 0; i < lines_out->nlines; i++)
+		lines[i].in = NULL;
+
+	lines->text = *text_out;
+	lines++;
+	for (i = 0; i + 1U < text_len;) {
+		if ((*text_out)[i++] == '\n') {
+			lines->text = &(*text_out)[i];
+			lines[-1].len = (size_t)(lines->text - lines[-1].text);
+			lines++;
+		}
+	}
+	lines[-1].len = (size_t)(&(*text_out)[text_len] - lines[-1].text);
+
+	return 0;
+
+fail:
+	free(*text_out);
+	*text_out = NULL;
+	return -1;
+}
+
+static enum successfulness
+rediff(const char *fname)
+{
+	struct text text_in, text_out;
+	char *text;
+	int fd, close_fd;
+	enum successfulness ret;
+
+	if (!strcmp(fname, "-")) {
+		fname = "<stdout>";
+		close_fd = 0;
+		fd = STDOUT_FILENO;
+		if (read_lines(&text_in, &text, STDIN_FILENO, "<stdin>"))
+			return ERROR;
+	} else {
+		close_fd = 1;
+		fd = open(fname, O_RDWR);
+		if (fd < 0) {
+			weprintf("open %s O_RDWR:", fname);
+			return ERROR;
+		}
+		if (read_lines(&text_in, &text, fd, "<stdin>")) {
+			close(fd);
+			return ERROR;
+		}
+		if (lseek(fd, 0, SEEK_SET) != 0) {
+			weprintf("lseek %s 0 SEEK_SET:", fname);
+			return ERROR;
+		}
+	}
+
+	ret = rediff_file(&text_out, &text_in, fname);
+	if (ret == ERROR) {
+		ret = ERROR;
+		goto out;
+	}
+
+	if (send_text(fd, fname, &text_out)) {
+		ret = ERROR;
+		goto out;
+	}
+
+	if (close_fd) {
+		off_t length = lseek(fd, 0, SEEK_CUR);
+		if (length < 0) {
+			weprintf("lseek %s 0 SEEK_CUR:", fname);
+			ret = ERROR;
+			goto out;
+		}
+		if (ftruncate(fd, length)) {
+			weprintf("ftruncate %s <current position>:", fname);
+			ret = ERROR;
+			goto out;
+		}
+	}
+
+out:
+	if (close_fd)
+		close(fd);
+	free(text_out.lines);
+	free(text_in.lines);
+	free(text);
+	return ret;
+}
+
+int
+main(int argc, char *argv[])
+{
+	enum successfulness ret = 0, r;
+
+	libsimple_default_failure_exit = 2;
+
+	ARGBEGIN {
+	default:
+		usage();
+	} ARGEND;
+
+	if (fstat(STDERR_FILENO, &(struct stat){0})) {
+		int fd;
+		if (errno != EBADF)
+			eprintf("fstat <stderr>:");
+		fd = open("/dev/null", O_WRONLY);
+		if (fd < 0)
+			eprintf("open /dev/null O_WRONLY:");
+		if (fd != STDERR_FILENO) {
+			if (dup2(fd, STDERR_FILENO) != STDERR_FILENO)
+				eprintf("dup2 /dev/null <stderr>:");
+			close(fd);
+		}
+	}
+
+	if (argc) {
+		for (; *argv; argv++) {
+			r = rediff(*argv);
+			ret = MAX(ret, r);
+		}
+	} else {
+		ret = rediff("-");
+	}
+
+	return (int)ret;
+}