1 ///////////////////////////////////////////////////////////////////////////////
2 //
3 /// \file       message.c
4 /// \brief      Printing messages
5 //
6 //  Author:     Lasse Collin
7 //
8 //  This file has been put into the public domain.
9 //  You can do whatever you want with this file.
10 //
11 ///////////////////////////////////////////////////////////////////////////////
12 
13 #include "private.h"
14 
15 #include <stdarg.h>
16 
17 
18 /// Number of the current file
19 static unsigned int files_pos = 0;
20 
21 /// Total number of input files; zero if unknown.
22 static unsigned int files_total;
23 
24 /// Verbosity level
25 static enum message_verbosity verbosity = V_WARNING;
26 
27 /// Filename which we will print with the verbose messages
28 static const char *filename;
29 
30 /// True once the a filename has been printed to stderr as part of progress
31 /// message. If automatic progress updating isn't enabled, this becomes true
32 /// after the first progress message has been printed due to user sending
33 /// SIGINFO, SIGUSR1, or SIGALRM. Once this variable is true, we will print
34 /// an empty line before the next filename to make the output more readable.
35 static bool first_filename_printed = false;
36 
37 /// This is set to true when we have printed the current filename to stderr
38 /// as part of a progress message. This variable is useful only if not
39 /// updating progress automatically: if user sends many SIGINFO, SIGUSR1, or
40 /// SIGALRM signals, we won't print the name of the same file multiple times.
41 static bool current_filename_printed = false;
42 
43 /// True if we should print progress indicator and update it automatically
44 /// if also verbose >= V_VERBOSE.
45 static bool progress_automatic;
46 
47 /// True if message_progress_start() has been called but
48 /// message_progress_end() hasn't been called yet.
49 static bool progress_started = false;
50 
51 /// This is true when a progress message was printed and the cursor is still
52 /// on the same line with the progress message. In that case, a newline has
53 /// to be printed before any error messages.
54 static bool progress_active = false;
55 
56 /// Pointer to lzma_stream used to do the encoding or decoding.
57 static lzma_stream *progress_strm;
58 
59 /// Expected size of the input stream is needed to show completion percentage
60 /// and estimate remaining time.
61 static uint64_t expected_in_size;
62 
63 
64 // Use alarm() and SIGALRM when they are supported. This has two minor
65 // advantages over the alternative of polling gettimeofday():
66 //  - It is possible for the user to send SIGINFO, SIGUSR1, or SIGALRM to
67 //    get intermediate progress information even when --verbose wasn't used
68 //    or stderr is not a terminal.
69 //  - alarm() + SIGALRM seems to have slightly less overhead than polling
70 //    gettimeofday().
71 #ifdef SIGALRM
72 
73 const int message_progress_sigs[] = {
74 	SIGALRM,
75 #ifdef SIGINFO
76 	SIGINFO,
77 #endif
78 #ifdef SIGUSR1
79 	SIGUSR1,
80 #endif
81 	0
82 };
83 
84 /// The signal handler for SIGALRM sets this to true. It is set back to false
85 /// once the progress message has been updated.
86 static volatile sig_atomic_t progress_needs_updating = false;
87 
88 /// Signal handler for SIGALRM
89 static void
progress_signal_handler(int sig lzma_attribute ((__unused__)))90 progress_signal_handler(int sig lzma_attribute((__unused__)))
91 {
92 	progress_needs_updating = true;
93 	return;
94 }
95 
96 #else
97 
98 /// This is true when progress message printing is wanted. Using the same
99 /// variable name as above to avoid some ifdefs.
100 static bool progress_needs_updating = false;
101 
102 /// Elapsed time when the next progress message update should be done.
103 static uint64_t progress_next_update;
104 
105 #endif
106 
107 
108 extern void
message_init(void)109 message_init(void)
110 {
111 	// If --verbose is used, we use a progress indicator if and only
112 	// if stderr is a terminal. If stderr is not a terminal, we print
113 	// verbose information only after finishing the file. As a special
114 	// exception, even if --verbose was not used, user can send SIGALRM
115 	// to make us print progress information once without automatic
116 	// updating.
117 	progress_automatic = isatty(STDERR_FILENO);
118 
119 	// Commented out because COLUMNS is rarely exported to environment.
120 	// Most users have at least 80 columns anyway, let's think something
121 	// fancy here if enough people complain.
122 /*
123 	if (progress_automatic) {
124 		// stderr is a terminal. Check the COLUMNS environment
125 		// variable to see if the terminal is wide enough. If COLUMNS
126 		// doesn't exist or it has some unparsable value, we assume
127 		// that the terminal is wide enough.
128 		const char *columns_str = getenv("COLUMNS");
129 		if (columns_str != NULL) {
130 			char *endptr;
131 			const long columns = strtol(columns_str, &endptr, 10);
132 			if (*endptr != '\0' || columns < 80)
133 				progress_automatic = false;
134 		}
135 	}
136 */
137 
138 #ifdef SIGALRM
139 	// Establish the signal handlers which set a flag to tell us that
140 	// progress info should be updated.
141 	struct sigaction sa;
142 	sigemptyset(&sa.sa_mask);
143 	sa.sa_flags = 0;
144 	sa.sa_handler = &progress_signal_handler;
145 
146 	for (size_t i = 0; message_progress_sigs[i] != 0; ++i)
147 		if (sigaction(message_progress_sigs[i], &sa, NULL))
148 			message_signal_handler();
149 #endif
150 
151 	return;
152 }
153 
154 
155 extern void
message_verbosity_increase(void)156 message_verbosity_increase(void)
157 {
158 	if (verbosity < V_DEBUG)
159 		++verbosity;
160 
161 	return;
162 }
163 
164 
165 extern void
message_verbosity_decrease(void)166 message_verbosity_decrease(void)
167 {
168 	if (verbosity > V_SILENT)
169 		--verbosity;
170 
171 	return;
172 }
173 
174 
175 extern enum message_verbosity
message_verbosity_get(void)176 message_verbosity_get(void)
177 {
178 	return verbosity;
179 }
180 
181 
182 extern void
message_set_files(unsigned int files)183 message_set_files(unsigned int files)
184 {
185 	files_total = files;
186 	return;
187 }
188 
189 
190 /// Prints the name of the current file if it hasn't been printed already,
191 /// except if we are processing exactly one stream from stdin to stdout.
192 /// I think it looks nicer to not print "(stdin)" when --verbose is used
193 /// in a pipe and no other files are processed.
194 static void
print_filename(void)195 print_filename(void)
196 {
197 	if (!opt_robot && (files_total != 1 || filename != stdin_filename)) {
198 		signals_block();
199 
200 		FILE *file = opt_mode == MODE_LIST ? stdout : stderr;
201 
202 		// If a file was already processed, put an empty line
203 		// before the next filename to improve readability.
204 		if (first_filename_printed)
205 			fputc('\n', file);
206 
207 		first_filename_printed = true;
208 		current_filename_printed = true;
209 
210 		// If we don't know how many files there will be due
211 		// to usage of --files or --files0.
212 		if (files_total == 0)
213 			fprintf(file, "%s (%u)\n", filename,
214 					files_pos);
215 		else
216 			fprintf(file, "%s (%u/%u)\n", filename,
217 					files_pos, files_total);
218 
219 		signals_unblock();
220 	}
221 
222 	return;
223 }
224 
225 
226 extern void
message_filename(const char * src_name)227 message_filename(const char *src_name)
228 {
229 	// Start numbering the files starting from one.
230 	++files_pos;
231 	filename = src_name;
232 
233 	if (verbosity >= V_VERBOSE
234 			&& (progress_automatic || opt_mode == MODE_LIST))
235 		print_filename();
236 	else
237 		current_filename_printed = false;
238 
239 	return;
240 }
241 
242 
243 extern void
message_progress_start(lzma_stream * strm,uint64_t in_size)244 message_progress_start(lzma_stream *strm, uint64_t in_size)
245 {
246 	// Store the pointer to the lzma_stream used to do the coding.
247 	// It is needed to find out the position in the stream.
248 	progress_strm = strm;
249 
250 	// Store the expected size of the file. If we aren't printing any
251 	// statistics, then is will be unused. But since it is possible
252 	// that the user sends us a signal to show statistics, we need
253 	// to have it available anyway.
254 	expected_in_size = in_size;
255 
256 	// Indicate that progress info may need to be printed before
257 	// printing error messages.
258 	progress_started = true;
259 
260 	// If progress indicator is wanted, print the filename and possibly
261 	// the file count now.
262 	if (verbosity >= V_VERBOSE && progress_automatic) {
263 		// Start the timer to display the first progress message
264 		// after one second. An alternative would be to show the
265 		// first message almost immediately, but delaying by one
266 		// second looks better to me, since extremely early
267 		// progress info is pretty much useless.
268 #ifdef SIGALRM
269 		// First disable a possibly existing alarm.
270 		alarm(0);
271 		progress_needs_updating = false;
272 		alarm(1);
273 #else
274 		progress_needs_updating = true;
275 		progress_next_update = 1000;
276 #endif
277 	}
278 
279 	return;
280 }
281 
282 
283 /// Make the string indicating completion percentage.
284 static const char *
progress_percentage(uint64_t in_pos)285 progress_percentage(uint64_t in_pos)
286 {
287 	// If the size of the input file is unknown or the size told us is
288 	// clearly wrong since we have processed more data than the alleged
289 	// size of the file, show a static string indicating that we have
290 	// no idea of the completion percentage.
291 	if (expected_in_size == 0 || in_pos > expected_in_size)
292 		return "--- %";
293 
294 	// Never show 100.0 % before we actually are finished.
295 	double percentage = (double)(in_pos) / (double)(expected_in_size)
296 			* 99.9;
297 
298 	// Use big enough buffer to hold e.g. a multibyte decimal point.
299 	static char buf[16];
300 	snprintf(buf, sizeof(buf), "%.1f %%", percentage);
301 
302 	return buf;
303 }
304 
305 
306 /// Make the string containing the amount of input processed, amount of
307 /// output produced, and the compression ratio.
308 static const char *
progress_sizes(uint64_t compressed_pos,uint64_t uncompressed_pos,bool final)309 progress_sizes(uint64_t compressed_pos, uint64_t uncompressed_pos, bool final)
310 {
311 	// Use big enough buffer to hold e.g. a multibyte thousand separators.
312 	static char buf[128];
313 	char *pos = buf;
314 	size_t left = sizeof(buf);
315 
316 	// Print the sizes. If this the final message, use more reasonable
317 	// units than MiB if the file was small.
318 	const enum nicestr_unit unit_min = final ? NICESTR_B : NICESTR_MIB;
319 	my_snprintf(&pos, &left, "%s / %s",
320 			uint64_to_nicestr(compressed_pos,
321 				unit_min, NICESTR_TIB, false, 0),
322 			uint64_to_nicestr(uncompressed_pos,
323 				unit_min, NICESTR_TIB, false, 1));
324 
325 	// Avoid division by zero. If we cannot calculate the ratio, set
326 	// it to some nice number greater than 10.0 so that it gets caught
327 	// in the next if-clause.
328 	const double ratio = uncompressed_pos > 0
329 			? (double)(compressed_pos) / (double)(uncompressed_pos)
330 			: 16.0;
331 
332 	// If the ratio is very bad, just indicate that it is greater than
333 	// 9.999. This way the length of the ratio field stays fixed.
334 	if (ratio > 9.999)
335 		snprintf(pos, left, " > %.3f", 9.999);
336 	else
337 		snprintf(pos, left, " = %.3f", ratio);
338 
339 	return buf;
340 }
341 
342 
343 /// Make the string containing the processing speed of uncompressed data.
344 static const char *
progress_speed(uint64_t uncompressed_pos,uint64_t elapsed)345 progress_speed(uint64_t uncompressed_pos, uint64_t elapsed)
346 {
347 	// Don't print the speed immediately, since the early values look
348 	// somewhat random.
349 	if (elapsed < 3000)
350 		return "";
351 
352 	static const char unit[][8] = {
353 		"KiB/s",
354 		"MiB/s",
355 		"GiB/s",
356 	};
357 
358 	size_t unit_index = 0;
359 
360 	// Calculate the speed as KiB/s.
361 	double speed = (double)(uncompressed_pos)
362 			/ ((double)(elapsed) * (1024.0 / 1000.0));
363 
364 	// Adjust the unit of the speed if needed.
365 	while (speed > 999.0) {
366 		speed /= 1024.0;
367 		if (++unit_index == ARRAY_SIZE(unit))
368 			return ""; // Way too fast ;-)
369 	}
370 
371 	// Use decimal point only if the number is small. Examples:
372 	//  - 0.1 KiB/s
373 	//  - 9.9 KiB/s
374 	//  - 99 KiB/s
375 	//  - 999 KiB/s
376 	// Use big enough buffer to hold e.g. a multibyte decimal point.
377 	static char buf[16];
378 	snprintf(buf, sizeof(buf), "%.*f %s",
379 			speed > 9.9 ? 0 : 1, speed, unit[unit_index]);
380 	return buf;
381 }
382 
383 
384 /// Make a string indicating elapsed or remaining time. The format is either
385 /// M:SS or H:MM:SS depending on if the time is an hour or more.
386 static const char *
progress_time(uint64_t mseconds)387 progress_time(uint64_t mseconds)
388 {
389 	// 9999 hours = 416 days
390 	static char buf[sizeof("9999:59:59")];
391 
392 	uint32_t seconds = mseconds / 1000;
393 
394 	// Don't show anything if the time is zero or ridiculously big.
395 	if (seconds == 0 || seconds > ((9999 * 60) + 59) * 60 + 59)
396 		return "";
397 
398 	uint32_t minutes = seconds / 60;
399 	seconds %= 60;
400 
401 	if (minutes >= 60) {
402 		const uint32_t hours = minutes / 60;
403 		minutes %= 60;
404 		snprintf(buf, sizeof(buf),
405 				"%" PRIu32 ":%02" PRIu32 ":%02" PRIu32,
406 				hours, minutes, seconds);
407 	} else {
408 		snprintf(buf, sizeof(buf), "%" PRIu32 ":%02" PRIu32,
409 				minutes, seconds);
410 	}
411 
412 	return buf;
413 }
414 
415 
416 /// Return a string containing estimated remaining time when
417 /// reasonably possible.
418 static const char *
progress_remaining(uint64_t in_pos,uint64_t elapsed)419 progress_remaining(uint64_t in_pos, uint64_t elapsed)
420 {
421 	// Don't show the estimated remaining time when it wouldn't
422 	// make sense:
423 	//  - Input size is unknown.
424 	//  - Input has grown bigger since we started (de)compressing.
425 	//  - We haven't processed much data yet, so estimate would be
426 	//    too inaccurate.
427 	//  - Only a few seconds has passed since we started (de)compressing,
428 	//    so estimate would be too inaccurate.
429 	if (expected_in_size == 0 || in_pos > expected_in_size
430 			|| in_pos < (UINT64_C(1) << 19) || elapsed < 8000)
431 		return "";
432 
433 	// Calculate the estimate. Don't give an estimate of zero seconds,
434 	// since it is possible that all the input has been already passed
435 	// to the library, but there is still quite a bit of output pending.
436 	uint32_t remaining = (double)(expected_in_size - in_pos)
437 			* ((double)(elapsed) / 1000.0) / (double)(in_pos);
438 	if (remaining < 1)
439 		remaining = 1;
440 
441 	static char buf[sizeof("9 h 55 min")];
442 
443 	// Select appropriate precision for the estimated remaining time.
444 	if (remaining <= 10) {
445 		// A maximum of 10 seconds remaining.
446 		// Show the number of seconds as is.
447 		snprintf(buf, sizeof(buf), "%" PRIu32 " s", remaining);
448 
449 	} else if (remaining <= 50) {
450 		// A maximum of 50 seconds remaining.
451 		// Round up to the next multiple of five seconds.
452 		remaining = (remaining + 4) / 5 * 5;
453 		snprintf(buf, sizeof(buf), "%" PRIu32 " s", remaining);
454 
455 	} else if (remaining <= 590) {
456 		// A maximum of 9 minutes and 50 seconds remaining.
457 		// Round up to the next multiple of ten seconds.
458 		remaining = (remaining + 9) / 10 * 10;
459 		snprintf(buf, sizeof(buf), "%" PRIu32 " min %" PRIu32 " s",
460 				remaining / 60, remaining % 60);
461 
462 	} else if (remaining <= 59 * 60) {
463 		// A maximum of 59 minutes remaining.
464 		// Round up to the next multiple of a minute.
465 		remaining = (remaining + 59) / 60;
466 		snprintf(buf, sizeof(buf), "%" PRIu32 " min", remaining);
467 
468 	} else if (remaining <= 9 * 3600 + 50 * 60) {
469 		// A maximum of 9 hours and 50 minutes left.
470 		// Round up to the next multiple of ten minutes.
471 		remaining = (remaining + 599) / 600 * 10;
472 		snprintf(buf, sizeof(buf), "%" PRIu32 " h %" PRIu32 " min",
473 				remaining / 60, remaining % 60);
474 
475 	} else if (remaining <= 23 * 3600) {
476 		// A maximum of 23 hours remaining.
477 		// Round up to the next multiple of an hour.
478 		remaining = (remaining + 3599) / 3600;
479 		snprintf(buf, sizeof(buf), "%" PRIu32 " h", remaining);
480 
481 	} else if (remaining <= 9 * 24 * 3600 + 23 * 3600) {
482 		// A maximum of 9 days and 23 hours remaining.
483 		// Round up to the next multiple of an hour.
484 		remaining = (remaining + 3599) / 3600;
485 		snprintf(buf, sizeof(buf), "%" PRIu32 " d %" PRIu32 " h",
486 				remaining / 24, remaining % 24);
487 
488 	} else if (remaining <= 999 * 24 * 3600) {
489 		// A maximum of 999 days remaining. ;-)
490 		// Round up to the next multiple of a day.
491 		remaining = (remaining + 24 * 3600 - 1) / (24 * 3600);
492 		snprintf(buf, sizeof(buf), "%" PRIu32 " d", remaining);
493 
494 	} else {
495 		// The estimated remaining time is too big. Don't show it.
496 		return "";
497 	}
498 
499 	return buf;
500 }
501 
502 
503 /// Get how much uncompressed and compressed data has been processed.
504 static void
progress_pos(uint64_t * in_pos,uint64_t * compressed_pos,uint64_t * uncompressed_pos)505 progress_pos(uint64_t *in_pos,
506 		uint64_t *compressed_pos, uint64_t *uncompressed_pos)
507 {
508 	uint64_t out_pos;
509 	lzma_get_progress(progress_strm, in_pos, &out_pos);
510 
511 	// It cannot have processed more input than it has been given.
512 	assert(*in_pos <= progress_strm->total_in);
513 
514 	// It cannot have produced more output than it claims to have ready.
515 	assert(out_pos >= progress_strm->total_out);
516 
517 	if (opt_mode == MODE_COMPRESS) {
518 		*compressed_pos = out_pos;
519 		*uncompressed_pos = *in_pos;
520 	} else {
521 		*compressed_pos = *in_pos;
522 		*uncompressed_pos = out_pos;
523 	}
524 
525 	return;
526 }
527 
528 
529 extern void
message_progress_update(void)530 message_progress_update(void)
531 {
532 	if (!progress_needs_updating)
533 		return;
534 
535 	// Calculate how long we have been processing this file.
536 	const uint64_t elapsed = mytime_get_elapsed();
537 
538 #ifndef SIGALRM
539 	if (progress_next_update > elapsed)
540 		return;
541 
542 	progress_next_update = elapsed + 1000;
543 #endif
544 
545 	// Get our current position in the stream.
546 	uint64_t in_pos;
547 	uint64_t compressed_pos;
548 	uint64_t uncompressed_pos;
549 	progress_pos(&in_pos, &compressed_pos, &uncompressed_pos);
550 
551 	// Block signals so that fprintf() doesn't get interrupted.
552 	signals_block();
553 
554 	// Print the filename if it hasn't been printed yet.
555 	if (!current_filename_printed)
556 		print_filename();
557 
558 	// Print the actual progress message. The idea is that there is at
559 	// least three spaces between the fields in typical situations, but
560 	// even in rare situations there is at least one space.
561 	const char *cols[5] = {
562 		progress_percentage(in_pos),
563 		progress_sizes(compressed_pos, uncompressed_pos, false),
564 		progress_speed(uncompressed_pos, elapsed),
565 		progress_time(elapsed),
566 		progress_remaining(in_pos, elapsed),
567 	};
568 	fprintf(stderr, "\r %*s %*s   %*s %10s   %10s\r",
569 			tuklib_mbstr_fw(cols[0], 6), cols[0],
570 			tuklib_mbstr_fw(cols[1], 35), cols[1],
571 			tuklib_mbstr_fw(cols[2], 9), cols[2],
572 			cols[3],
573 			cols[4]);
574 
575 #ifdef SIGALRM
576 	// Updating the progress info was finished. Reset
577 	// progress_needs_updating to wait for the next SIGALRM.
578 	//
579 	// NOTE: This has to be done before alarm(1) or with (very) bad
580 	// luck we could be setting this to false after the alarm has already
581 	// been triggered.
582 	progress_needs_updating = false;
583 
584 	if (verbosity >= V_VERBOSE && progress_automatic) {
585 		// Mark that the progress indicator is active, so if an error
586 		// occurs, the error message gets printed cleanly.
587 		progress_active = true;
588 
589 		// Restart the timer so that progress_needs_updating gets
590 		// set to true after about one second.
591 		alarm(1);
592 	} else {
593 		// The progress message was printed because user had sent us
594 		// SIGALRM. In this case, each progress message is printed
595 		// on its own line.
596 		fputc('\n', stderr);
597 	}
598 #else
599 	// When SIGALRM isn't supported and we get here, it's always due to
600 	// automatic progress update. We set progress_active here too like
601 	// described above.
602 	assert(verbosity >= V_VERBOSE);
603 	assert(progress_automatic);
604 	progress_active = true;
605 #endif
606 
607 	signals_unblock();
608 
609 	return;
610 }
611 
612 
613 static void
progress_flush(bool finished)614 progress_flush(bool finished)
615 {
616 	if (!progress_started || verbosity < V_VERBOSE)
617 		return;
618 
619 	uint64_t in_pos;
620 	uint64_t compressed_pos;
621 	uint64_t uncompressed_pos;
622 	progress_pos(&in_pos, &compressed_pos, &uncompressed_pos);
623 
624 	// Avoid printing intermediate progress info if some error occurs
625 	// in the beginning of the stream. (If something goes wrong later in
626 	// the stream, it is sometimes useful to tell the user where the
627 	// error approximately occurred, especially if the error occurs
628 	// after a time-consuming operation.)
629 	if (!finished && !progress_active
630 			&& (compressed_pos == 0 || uncompressed_pos == 0))
631 		return;
632 
633 	progress_active = false;
634 
635 	const uint64_t elapsed = mytime_get_elapsed();
636 
637 	signals_block();
638 
639 	// When using the auto-updating progress indicator, the final
640 	// statistics are printed in the same format as the progress
641 	// indicator itself.
642 	if (progress_automatic) {
643 		const char *cols[5] = {
644 			finished ? "100 %" : progress_percentage(in_pos),
645 			progress_sizes(compressed_pos, uncompressed_pos, true),
646 			progress_speed(uncompressed_pos, elapsed),
647 			progress_time(elapsed),
648 			finished ? "" : progress_remaining(in_pos, elapsed),
649 		};
650 		fprintf(stderr, "\r %*s %*s   %*s %10s   %10s\n",
651 				tuklib_mbstr_fw(cols[0], 6), cols[0],
652 				tuklib_mbstr_fw(cols[1], 35), cols[1],
653 				tuklib_mbstr_fw(cols[2], 9), cols[2],
654 				cols[3],
655 				cols[4]);
656 	} else {
657 		// The filename is always printed.
658 		fprintf(stderr, "%s: ", filename);
659 
660 		// Percentage is printed only if we didn't finish yet.
661 		if (!finished) {
662 			// Don't print the percentage when it isn't known
663 			// (starts with a dash).
664 			const char *percentage = progress_percentage(in_pos);
665 			if (percentage[0] != '-')
666 				fprintf(stderr, "%s, ", percentage);
667 		}
668 
669 		// Size information is always printed.
670 		fprintf(stderr, "%s", progress_sizes(
671 				compressed_pos, uncompressed_pos, true));
672 
673 		// The speed and elapsed time aren't always shown.
674 		const char *speed = progress_speed(uncompressed_pos, elapsed);
675 		if (speed[0] != '\0')
676 			fprintf(stderr, ", %s", speed);
677 
678 		const char *elapsed_str = progress_time(elapsed);
679 		if (elapsed_str[0] != '\0')
680 			fprintf(stderr, ", %s", elapsed_str);
681 
682 		fputc('\n', stderr);
683 	}
684 
685 	signals_unblock();
686 
687 	return;
688 }
689 
690 
691 extern void
message_progress_end(bool success)692 message_progress_end(bool success)
693 {
694 	assert(progress_started);
695 	progress_flush(success);
696 	progress_started = false;
697 	return;
698 }
699 
700 
701 static void
vmessage(enum message_verbosity v,const char * fmt,va_list ap)702 vmessage(enum message_verbosity v, const char *fmt, va_list ap)
703 {
704 	if (v <= verbosity) {
705 		signals_block();
706 
707 		progress_flush(false);
708 
709 		// TRANSLATORS: This is the program name in the beginning
710 		// of the line in messages. Usually it becomes "xz: ".
711 		// This is a translatable string because French needs
712 		// a space before a colon.
713 		fprintf(stderr, _("%s: "), progname);
714 		vfprintf(stderr, fmt, ap);
715 		fputc('\n', stderr);
716 
717 		signals_unblock();
718 	}
719 
720 	return;
721 }
722 
723 
724 extern void
message(enum message_verbosity v,const char * fmt,...)725 message(enum message_verbosity v, const char *fmt, ...)
726 {
727 	va_list ap;
728 	va_start(ap, fmt);
729 	vmessage(v, fmt, ap);
730 	va_end(ap);
731 	return;
732 }
733 
734 
735 extern void
message_warning(const char * fmt,...)736 message_warning(const char *fmt, ...)
737 {
738 	va_list ap;
739 	va_start(ap, fmt);
740 	vmessage(V_WARNING, fmt, ap);
741 	va_end(ap);
742 
743 	set_exit_status(E_WARNING);
744 	return;
745 }
746 
747 
748 extern void
message_error(const char * fmt,...)749 message_error(const char *fmt, ...)
750 {
751 	va_list ap;
752 	va_start(ap, fmt);
753 	vmessage(V_ERROR, fmt, ap);
754 	va_end(ap);
755 
756 	set_exit_status(E_ERROR);
757 	return;
758 }
759 
760 
761 extern void
message_fatal(const char * fmt,...)762 message_fatal(const char *fmt, ...)
763 {
764 	va_list ap;
765 	va_start(ap, fmt);
766 	vmessage(V_ERROR, fmt, ap);
767 	va_end(ap);
768 
769 	tuklib_exit(E_ERROR, E_ERROR, false);
770 }
771 
772 
773 extern void
message_bug(void)774 message_bug(void)
775 {
776 	message_fatal(_("Internal error (bug)"));
777 }
778 
779 
780 extern void
message_signal_handler(void)781 message_signal_handler(void)
782 {
783 	message_fatal(_("Cannot establish signal handlers"));
784 }
785 
786 
787 extern const char *
message_strm(lzma_ret code)788 message_strm(lzma_ret code)
789 {
790 	switch (code) {
791 	case LZMA_NO_CHECK:
792 		return _("No integrity check; not verifying file integrity");
793 
794 	case LZMA_UNSUPPORTED_CHECK:
795 		return _("Unsupported type of integrity check; "
796 				"not verifying file integrity");
797 
798 	case LZMA_MEM_ERROR:
799 		return strerror(ENOMEM);
800 
801 	case LZMA_MEMLIMIT_ERROR:
802 		return _("Memory usage limit reached");
803 
804 	case LZMA_FORMAT_ERROR:
805 		return _("File format not recognized");
806 
807 	case LZMA_OPTIONS_ERROR:
808 		return _("Unsupported options");
809 
810 	case LZMA_DATA_ERROR:
811 		return _("Compressed data is corrupt");
812 
813 	case LZMA_BUF_ERROR:
814 		return _("Unexpected end of input");
815 
816 	case LZMA_OK:
817 	case LZMA_STREAM_END:
818 	case LZMA_GET_CHECK:
819 	case LZMA_PROG_ERROR:
820 		// Without "default", compiler will warn if new constants
821 		// are added to lzma_ret, it is not too easy to forget to
822 		// add the new constants to this function.
823 		break;
824 	}
825 
826 	return _("Internal error (bug)");
827 }
828 
829 
830 extern void
message_mem_needed(enum message_verbosity v,uint64_t memusage)831 message_mem_needed(enum message_verbosity v, uint64_t memusage)
832 {
833 	if (v > verbosity)
834 		return;
835 
836 	// Convert memusage to MiB, rounding up to the next full MiB.
837 	// This way the user can always use the displayed usage as
838 	// the new memory usage limit. (If we rounded to the nearest,
839 	// the user might need to +1 MiB to get high enough limit.)
840 	memusage = round_up_to_mib(memusage);
841 
842 	uint64_t memlimit = hardware_memlimit_get(opt_mode);
843 
844 	// Handle the case when there is no memory usage limit.
845 	// This way we don't print a weird message with a huge number.
846 	if (memlimit == UINT64_MAX) {
847 		message(v, _("%s MiB of memory is required. "
848 				"The limiter is disabled."),
849 				uint64_to_str(memusage, 0));
850 		return;
851 	}
852 
853 	// With US-ASCII:
854 	// 2^64 with thousand separators + " MiB" suffix + '\0' = 26 + 4 + 1
855 	// But there may be multibyte chars so reserve enough space.
856 	char memlimitstr[128];
857 
858 	// Show the memory usage limit as MiB unless it is less than 1 MiB.
859 	// This way it's easy to notice errors where one has typed
860 	// --memory=123 instead of --memory=123MiB.
861 	if (memlimit < (UINT32_C(1) << 20)) {
862 		snprintf(memlimitstr, sizeof(memlimitstr), "%s B",
863 				uint64_to_str(memlimit, 1));
864 	} else {
865 		// Round up just like with memusage. If this function is
866 		// called for informational purposes (to just show the
867 		// current usage and limit), we should never show that
868 		// the usage is higher than the limit, which would give
869 		// a false impression that the memory usage limit isn't
870 		// properly enforced.
871 		snprintf(memlimitstr, sizeof(memlimitstr), "%s MiB",
872 				uint64_to_str(round_up_to_mib(memlimit), 1));
873 	}
874 
875 	message(v, _("%s MiB of memory is required. The limit is %s."),
876 			uint64_to_str(memusage, 0), memlimitstr);
877 
878 	return;
879 }
880 
881 
882 /// \brief      Convert uint32_t to a nice string for --lzma[12]=dict=SIZE
883 ///
884 /// The idea is to use KiB or MiB suffix when possible.
885 static const char *
uint32_to_optstr(uint32_t num)886 uint32_to_optstr(uint32_t num)
887 {
888 	static char buf[16];
889 
890 	if ((num & ((UINT32_C(1) << 20) - 1)) == 0)
891 		snprintf(buf, sizeof(buf), "%" PRIu32 "MiB", num >> 20);
892 	else if ((num & ((UINT32_C(1) << 10) - 1)) == 0)
893 		snprintf(buf, sizeof(buf), "%" PRIu32 "KiB", num >> 10);
894 	else
895 		snprintf(buf, sizeof(buf), "%" PRIu32, num);
896 
897 	return buf;
898 }
899 
900 
901 extern void
message_filters_to_str(char buf[FILTERS_STR_SIZE],const lzma_filter * filters,bool all_known)902 message_filters_to_str(char buf[FILTERS_STR_SIZE],
903 		const lzma_filter *filters, bool all_known)
904 {
905 	char *pos = buf;
906 	size_t left = FILTERS_STR_SIZE;
907 
908 	for (size_t i = 0; filters[i].id != LZMA_VLI_UNKNOWN; ++i) {
909 		// Add the dashes for the filter option. A space is
910 		// needed after the first and later filters.
911 		my_snprintf(&pos, &left, "%s", i == 0 ? "--" : " --");
912 
913 		switch (filters[i].id) {
914 		case LZMA_FILTER_LZMA1:
915 		case LZMA_FILTER_LZMA2: {
916 			const lzma_options_lzma *opt = filters[i].options;
917 			const char *mode = NULL;
918 			const char *mf = NULL;
919 
920 			if (all_known) {
921 				switch (opt->mode) {
922 				case LZMA_MODE_FAST:
923 					mode = "fast";
924 					break;
925 
926 				case LZMA_MODE_NORMAL:
927 					mode = "normal";
928 					break;
929 
930 				default:
931 					mode = "UNKNOWN";
932 					break;
933 				}
934 
935 				switch (opt->mf) {
936 				case LZMA_MF_HC3:
937 					mf = "hc3";
938 					break;
939 
940 				case LZMA_MF_HC4:
941 					mf = "hc4";
942 					break;
943 
944 				case LZMA_MF_BT2:
945 					mf = "bt2";
946 					break;
947 
948 				case LZMA_MF_BT3:
949 					mf = "bt3";
950 					break;
951 
952 				case LZMA_MF_BT4:
953 					mf = "bt4";
954 					break;
955 
956 				default:
957 					mf = "UNKNOWN";
958 					break;
959 				}
960 			}
961 
962 			// Add the filter name and dictionary size, which
963 			// is always known.
964 			my_snprintf(&pos, &left, "lzma%c=dict=%s",
965 					filters[i].id == LZMA_FILTER_LZMA2
966 						? '2' : '1',
967 					uint32_to_optstr(opt->dict_size));
968 
969 			// With LZMA1 also lc/lp/pb are known when
970 			// decompressing, but this function is never
971 			// used to print information about .lzma headers.
972 			assert(filters[i].id == LZMA_FILTER_LZMA2
973 					|| all_known);
974 
975 			// Print the rest of the options, which are known
976 			// only when compressing.
977 			if (all_known)
978 				my_snprintf(&pos, &left,
979 					",lc=%" PRIu32 ",lp=%" PRIu32
980 					",pb=%" PRIu32
981 					",mode=%s,nice=%" PRIu32 ",mf=%s"
982 					",depth=%" PRIu32,
983 					opt->lc, opt->lp, opt->pb,
984 					mode, opt->nice_len, mf, opt->depth);
985 			break;
986 		}
987 
988 		case LZMA_FILTER_X86:
989 		case LZMA_FILTER_POWERPC:
990 		case LZMA_FILTER_IA64:
991 		case LZMA_FILTER_ARM:
992 		case LZMA_FILTER_ARMTHUMB:
993 		case LZMA_FILTER_SPARC: {
994 			static const char bcj_names[][9] = {
995 				"x86",
996 				"powerpc",
997 				"ia64",
998 				"arm",
999 				"armthumb",
1000 				"sparc",
1001 			};
1002 
1003 			const lzma_options_bcj *opt = filters[i].options;
1004 			my_snprintf(&pos, &left, "%s", bcj_names[filters[i].id
1005 					- LZMA_FILTER_X86]);
1006 
1007 			// Show the start offset only when really needed.
1008 			if (opt != NULL && opt->start_offset != 0)
1009 				my_snprintf(&pos, &left, "=start=%" PRIu32,
1010 						opt->start_offset);
1011 
1012 			break;
1013 		}
1014 
1015 		case LZMA_FILTER_DELTA: {
1016 			const lzma_options_delta *opt = filters[i].options;
1017 			my_snprintf(&pos, &left, "delta=dist=%" PRIu32,
1018 					opt->dist);
1019 			break;
1020 		}
1021 
1022 		default:
1023 			// This should be possible only if liblzma is
1024 			// newer than the xz tool.
1025 			my_snprintf(&pos, &left, "UNKNOWN");
1026 			break;
1027 		}
1028 	}
1029 
1030 	return;
1031 }
1032 
1033 
1034 extern void
message_filters_show(enum message_verbosity v,const lzma_filter * filters)1035 message_filters_show(enum message_verbosity v, const lzma_filter *filters)
1036 {
1037 	if (v > verbosity)
1038 		return;
1039 
1040 	char buf[FILTERS_STR_SIZE];
1041 	message_filters_to_str(buf, filters, true);
1042 	fprintf(stderr, _("%s: Filter chain: %s\n"), progname, buf);
1043 	return;
1044 }
1045 
1046 
1047 extern void
message_try_help(void)1048 message_try_help(void)
1049 {
1050 	// Print this with V_WARNING instead of V_ERROR to prevent it from
1051 	// showing up when --quiet has been specified.
1052 	message(V_WARNING, _("Try `%s --help' for more information."),
1053 			progname);
1054 	return;
1055 }
1056 
1057 
1058 extern void
message_version(void)1059 message_version(void)
1060 {
1061 	// It is possible that liblzma version is different than the command
1062 	// line tool version, so print both.
1063 	if (opt_robot) {
1064 		printf("XZ_VERSION=%" PRIu32 "\nLIBLZMA_VERSION=%" PRIu32 "\n",
1065 				LZMA_VERSION, lzma_version_number());
1066 	} else {
1067 		printf("xz (" PACKAGE_NAME ") " LZMA_VERSION_STRING "\n");
1068 		printf("liblzma %s\n", lzma_version_string());
1069 	}
1070 
1071 	tuklib_exit(E_SUCCESS, E_ERROR, verbosity != V_SILENT);
1072 }
1073 
1074 
1075 extern void
message_help(bool long_help)1076 message_help(bool long_help)
1077 {
1078 	printf(_("Usage: %s [OPTION]... [FILE]...\n"
1079 			"Compress or decompress FILEs in the .xz format.\n\n"),
1080 			progname);
1081 
1082 	// NOTE: The short help doesn't currently have options that
1083 	// take arguments.
1084 	if (long_help)
1085 		puts(_("Mandatory arguments to long options are mandatory "
1086 				"for short options too.\n"));
1087 
1088 	if (long_help)
1089 		puts(_(" Operation mode:\n"));
1090 
1091 	puts(_(
1092 "  -z, --compress      force compression\n"
1093 "  -d, --decompress    force decompression\n"
1094 "  -t, --test          test compressed file integrity\n"
1095 "  -l, --list          list information about .xz files"));
1096 
1097 	if (long_help)
1098 		puts(_("\n Operation modifiers:\n"));
1099 
1100 	puts(_(
1101 "  -k, --keep          keep (don't delete) input files\n"
1102 "  -f, --force         force overwrite of output file and (de)compress links\n"
1103 "  -c, --stdout        write to standard output and don't delete input files"));
1104 
1105 	if (long_help) {
1106 		puts(_(
1107 "      --single-stream decompress only the first stream, and silently\n"
1108 "                      ignore possible remaining input data"));
1109 		puts(_(
1110 "      --no-sparse     do not create sparse files when decompressing\n"
1111 "  -S, --suffix=.SUF   use the suffix `.SUF' on compressed files\n"
1112 "      --files[=FILE]  read filenames to process from FILE; if FILE is\n"
1113 "                      omitted, filenames are read from the standard input;\n"
1114 "                      filenames must be terminated with the newline character\n"
1115 "      --files0[=FILE] like --files but use the null character as terminator"));
1116 	}
1117 
1118 	if (long_help) {
1119 		puts(_("\n Basic file format and compression options:\n"));
1120 		puts(_(
1121 "  -F, --format=FMT    file format to encode or decode; possible values are\n"
1122 "                      `auto' (default), `xz', `lzma', and `raw'\n"
1123 "  -C, --check=CHECK   integrity check type: `none' (use with caution),\n"
1124 "                      `crc32', `crc64' (default), or `sha256'"));
1125 		puts(_(
1126 "      --ignore-check  don't verify the integrity check when decompressing"));
1127 	}
1128 
1129 	puts(_(
1130 "  -0 ... -9           compression preset; default is 6; take compressor *and*\n"
1131 "                      decompressor memory usage into account before using 7-9!"));
1132 
1133 	puts(_(
1134 "  -e, --extreme       try to improve compression ratio by using more CPU time;\n"
1135 "                      does not affect decompressor memory requirements"));
1136 
1137 	puts(_(
1138 "  -T, --threads=NUM   use at most NUM threads; the default is 1; set to 0\n"
1139 "                      to use as many threads as there are processor cores"));
1140 
1141 	if (long_help) {
1142 		puts(_(
1143 "      --block-size=SIZE\n"
1144 "                      start a new .xz block after every SIZE bytes of input;\n"
1145 "                      use this to set the block size for threaded compression"));
1146 		puts(_(
1147 "      --block-list=SIZES\n"
1148 "                      start a new .xz block after the given comma-separated\n"
1149 "                      intervals of uncompressed data"));
1150 		puts(_(
1151 "      --flush-timeout=TIMEOUT\n"
1152 "                      when compressing, if more than TIMEOUT milliseconds has\n"
1153 "                      passed since the previous flush and reading more input\n"
1154 "                      would block, all pending data is flushed out"
1155 		));
1156 		puts(_( // xgettext:no-c-format
1157 "      --memlimit-compress=LIMIT\n"
1158 "      --memlimit-decompress=LIMIT\n"
1159 "  -M, --memlimit=LIMIT\n"
1160 "                      set memory usage limit for compression, decompression,\n"
1161 "                      or both; LIMIT is in bytes, % of RAM, or 0 for defaults"));
1162 
1163 		puts(_(
1164 "      --no-adjust     if compression settings exceed the memory usage limit,\n"
1165 "                      give an error instead of adjusting the settings downwards"));
1166 	}
1167 
1168 	if (long_help) {
1169 		puts(_(
1170 "\n Custom filter chain for compression (alternative for using presets):"));
1171 
1172 #if defined(HAVE_ENCODER_LZMA1) || defined(HAVE_DECODER_LZMA1) \
1173 		|| defined(HAVE_ENCODER_LZMA2) || defined(HAVE_DECODER_LZMA2)
1174 		// TRANSLATORS: The word "literal" in "literal context bits"
1175 		// means how many "context bits" to use when encoding
1176 		// literals. A literal is a single 8-bit byte. It doesn't
1177 		// mean "literally" here.
1178 		puts(_(
1179 "\n"
1180 "  --lzma1[=OPTS]      LZMA1 or LZMA2; OPTS is a comma-separated list of zero or\n"
1181 "  --lzma2[=OPTS]      more of the following options (valid values; default):\n"
1182 "                        preset=PRE reset options to a preset (0-9[e])\n"
1183 "                        dict=NUM   dictionary size (4KiB - 1536MiB; 8MiB)\n"
1184 "                        lc=NUM     number of literal context bits (0-4; 3)\n"
1185 "                        lp=NUM     number of literal position bits (0-4; 0)\n"
1186 "                        pb=NUM     number of position bits (0-4; 2)\n"
1187 "                        mode=MODE  compression mode (fast, normal; normal)\n"
1188 "                        nice=NUM   nice length of a match (2-273; 64)\n"
1189 "                        mf=NAME    match finder (hc3, hc4, bt2, bt3, bt4; bt4)\n"
1190 "                        depth=NUM  maximum search depth; 0=automatic (default)"));
1191 #endif
1192 
1193 		puts(_(
1194 "\n"
1195 "  --x86[=OPTS]        x86 BCJ filter (32-bit and 64-bit)\n"
1196 "  --powerpc[=OPTS]    PowerPC BCJ filter (big endian only)\n"
1197 "  --ia64[=OPTS]       IA-64 (Itanium) BCJ filter\n"
1198 "  --arm[=OPTS]        ARM BCJ filter (little endian only)\n"
1199 "  --armthumb[=OPTS]   ARM-Thumb BCJ filter (little endian only)\n"
1200 "  --sparc[=OPTS]      SPARC BCJ filter\n"
1201 "                      Valid OPTS for all BCJ filters:\n"
1202 "                        start=NUM  start offset for conversions (default=0)"));
1203 
1204 #if defined(HAVE_ENCODER_DELTA) || defined(HAVE_DECODER_DELTA)
1205 		puts(_(
1206 "\n"
1207 "  --delta[=OPTS]      Delta filter; valid OPTS (valid values; default):\n"
1208 "                        dist=NUM   distance between bytes being subtracted\n"
1209 "                                   from each other (1-256; 1)"));
1210 #endif
1211 	}
1212 
1213 	if (long_help)
1214 		puts(_("\n Other options:\n"));
1215 
1216 	puts(_(
1217 "  -q, --quiet         suppress warnings; specify twice to suppress errors too\n"
1218 "  -v, --verbose       be verbose; specify twice for even more verbose"));
1219 
1220 	if (long_help) {
1221 		puts(_(
1222 "  -Q, --no-warn       make warnings not affect the exit status"));
1223 		puts(_(
1224 "      --robot         use machine-parsable messages (useful for scripts)"));
1225 		puts("");
1226 		puts(_(
1227 "      --info-memory   display the total amount of RAM and the currently active\n"
1228 "                      memory usage limits, and exit"));
1229 		puts(_(
1230 "  -h, --help          display the short help (lists only the basic options)\n"
1231 "  -H, --long-help     display this long help and exit"));
1232 	} else {
1233 		puts(_(
1234 "  -h, --help          display this short help and exit\n"
1235 "  -H, --long-help     display the long help (lists also the advanced options)"));
1236 	}
1237 
1238 	puts(_(
1239 "  -V, --version       display the version number and exit"));
1240 
1241 	puts(_("\nWith no FILE, or when FILE is -, read standard input.\n"));
1242 
1243 	// TRANSLATORS: This message indicates the bug reporting address
1244 	// for this package. Please add _another line_ saying
1245 	// "Report translation bugs to <...>\n" with the email or WWW
1246 	// address for translation bugs. Thanks.
1247 	printf(_("Report bugs to <%s> (in English or Finnish).\n"),
1248 			PACKAGE_BUGREPORT);
1249 	printf(_("%s home page: <%s>\n"), PACKAGE_NAME, PACKAGE_URL);
1250 
1251 #if LZMA_VERSION_STABILITY != LZMA_VERSION_STABILITY_STABLE
1252 	puts(_(
1253 "THIS IS A DEVELOPMENT VERSION NOT INTENDED FOR PRODUCTION USE."));
1254 #endif
1255 
1256 	tuklib_exit(E_SUCCESS, E_ERROR, verbosity != V_SILENT);
1257 }
1258