1GCOV(1) GNU GCOV(1)
2
6 gcov - coverage testing tool
7
9 gcov [-v|--version] [-h|--help]
10 [-a|--all-blocks]
11 [-b|--branch-probabilities]
12 [-c|--branch-counts]
13 [-d|--display-progress]
14 [-f|--function-summaries]
15 [-i|--intermediate-format]
16 [-j|--human-readable]
17 [-k|--use-colors]
18 [-l|--long-file-names]
19 [-m|--demangled-names]
20 [-n|--no-output]
21 [-o|--object-directory directory|file]
22 [-p|--preserve-paths]
23 [-r|--relative-only]
24 [-s|--source-prefix directory]
25 [-u|--unconditional-branches]
26 [-x|--hash-filenames]
27 files
28
30 gcov is a test coverage program. Use it in concert with GCC to analyze
31 your programs to help create more efficient, faster running code and to
32 discover untested parts of your program. You can use gcov as a
33 profiling tool to help discover where your optimization efforts will
34 best affect your code. You can also use gcov along with the other
35 profiling tool, gprof, to assess which parts of your code use the
36 greatest amount of computing time.
37 Profiling tools help you analyze your code's performance. Using a
39 profiler such as gcov or gprof, you can find out some basic performance
40 statistics, such as:
41 * how often each line of code executes
43 * what lines of code are actually executed
45 * how much computing time each section of code uses
47 Once you know these things about how your code works when compiled, you
49 can look at each module to see which modules should be optimized. gcov
50 helps you determine where to work on optimization.
51 Software developers also use coverage testing in concert with
53 testsuites, to make sure software is actually good enough for a
54 release. Testsuites can verify that a program works as expected; a
55 coverage program tests to see how much of the program is exercised by
56 the testsuite. Developers can then determine what kinds of test cases
57 need to be added to the testsuites to create both better testing and a
58 better final product.
59 You should compile your code without optimization if you plan to use
61 gcov because the optimization, by combining some lines of code into one
62 function, may not give you as much information as you need to look for
63 `hot spots' where the code is using a great deal of computer time.
64 Likewise, because gcov accumulates statistics by line (at the lowest
65 resolution), it works best with a programming style that places only
66 one statement on each line. If you use complicated macros that expand
67 to loops or to other control structures, the statistics are less
68 helpful---they only report on the line where the macro call appears.
69 If your complex macros behave like functions, you can replace them with
70 inline functions to solve this problem.
71 gcov creates a logfile called sourcefile.gcov which indicates how many
73 times each line of a source file sourcefile.c has executed. You can
74 use these logfiles along with gprof to aid in fine-tuning the
75 performance of your programs. gprof gives timing information you can
76 use along with the information you get from gcov.
77 gcov works only on code compiled with GCC. It is not compatible with
79 any other profiling or test coverage mechanism.
80
82 -a
83 --all-blocks
84 Write individual execution counts for every basic block. Normally
85 gcov outputs execution counts only for the main blocks of a line.
86 With this option you can determine if blocks within a single line
87 are not being executed.
88 -b
90 --branch-probabilities
91 Write branch frequencies to the output file, and write branch
92 summary info to the standard output. This option allows you to see
93 how often each branch in your program was taken. Unconditional
94 branches will not be shown, unless the -u option is given.
95 -c
97 --branch-counts
98 Write branch frequencies as the number of branches taken, rather
99 than the percentage of branches taken.
100 -d
102 --display-progress
103 Display the progress on the standard output.
104 -f
106 --function-summaries
107 Output summaries for each function in addition to the file level
108 summary.
109 -h
111 --help
112 Display help about using gcov (on the standard output), and exit
113 without doing any further processing.
114 -i
116 --intermediate-format
117 Output gcov file in an easy-to-parse intermediate text format that
118 can be used by lcov or other tools. The output is a single .gcov
119 file per .gcda file. No source code is required.
120 The format of the intermediate .gcov file is plain text with one
122 entry per line
123 version:<gcc_version>
125 file:<source_file_name>
126 function:<start_line_number>,<end_line_number>,<execution_count>,<function_name>
127 lcount:<line number>,<execution_count>,<has_unexecuted_block>
128 branch:<line_number>,<branch_coverage_type>
129 Where the <branch_coverage_type> is
131 notexec (Branch not executed)
132 taken (Branch executed and taken)
133 nottaken (Branch executed, but not taken)
134 There can be multiple file entries in an intermediate gcov file.
136 All entries following a file pertain to that source file until the
137 next file entry. If there are multiple functions that start on a
138 single line, then corresponding lcount is repeated multiple times.
139 Here is a sample when -i is used in conjunction with -b option:
141 version: 8.1.0 20180103
143 file:tmp.cpp
144 function:7,7,0,_ZN3FooIcEC2Ev
145 function:7,7,1,_ZN3FooIiEC2Ev
146 function:8,8,0,_ZN3FooIcE3incEv
147 function:8,8,2,_ZN3FooIiE3incEv
148 function:18,37,1,main
149 lcount:7,0,1
150 lcount:7,1,0
151 lcount:8,0,1
152 lcount:8,2,0
153 lcount:18,1,0
154 lcount:21,1,0
155 branch:21,taken
156 branch:21,nottaken
157 lcount:23,1,0
158 branch:23,taken
159 branch:23,nottaken
160 lcount:24,1,0
161 branch:24,taken
162 branch:24,nottaken
163 lcount:25,1,0
164 lcount:27,11,0
165 branch:27,taken
166 branch:27,taken
167 lcount:28,10,0
168 lcount:30,1,1
169 branch:30,nottaken
170 branch:30,taken
171 lcount:32,1,0
172 branch:32,nottaken
173 branch:32,taken
174 lcount:33,0,1
175 branch:33,notexec
176 branch:33,notexec
177 lcount:35,1,0
178 branch:35,taken
179 branch:35,nottaken
180 lcount:36,1,0
181 -j
183 --human-readable
184 Write counts in human readable format (like 24k).
185 -k
187 --use-colors
188 Use colors for lines of code that have zero coverage. We use red
189 color for non-exceptional lines and cyan for exceptional. Same
190 colors are used for basic blocks with -a option.
191 -l
193 --long-file-names
194 Create long file names for included source files. For example, if
195 the header file x.h contains code, and was included in the file
196 a.c, then running gcov on the file a.c will produce an output file
197 called a.c##x.h.gcov instead of x.h.gcov. This can be useful if
198 x.h is included in multiple source files and you want to see the
199 individual contributions. If you use the -p option, both the
200 including and included file names will be complete path names.
201 -m
203 --demangled-names
204 Display demangled function names in output. The default is to show
205 mangled function names.
206 -n
208 --no-output
209 Do not create the gcov output file.
210 -o directory|file
212 --object-directory directory
213 --object-file file
214 Specify either the directory containing the gcov data files, or the
215 object path name. The .gcno, and .gcda data files are searched for
216 using this option. If a directory is specified, the data files are
217 in that directory and named after the input file name, without its
218 extension. If a file is specified here, the data files are named
219 after that file, without its extension.
220 -p
222 --preserve-paths
223 Preserve complete path information in the names of generated .gcov
224 files. Without this option, just the filename component is used.
225 With this option, all directories are used, with / characters
226 translated to # characters, . directory components removed and
227 unremoveable .. components renamed to ^. This is useful if
228 sourcefiles are in several different directories.
229 -r
231 --relative-only
232 Only output information about source files with a relative pathname
233 (after source prefix elision). Absolute paths are usually system
234 header files and coverage of any inline functions therein is
235 normally uninteresting.
236 -s directory
238 --source-prefix directory
239 A prefix for source file names to remove when generating the output
240 coverage files. This option is useful when building in a separate
241 directory, and the pathname to the source directory is not wanted
242 when determining the output file names. Note that this prefix
243 detection is applied before determining whether the source file is
244 absolute.
245 -u
247 --unconditional-branches
248 When branch probabilities are given, include those of unconditional
249 branches. Unconditional branches are normally not interesting.
250 -v
252 --version
253 Display the gcov version number (on the standard output), and exit
254 without doing any further processing.
255 -w
257 --verbose
258 Print verbose informations related to basic blocks and arcs.
259 -x
261 --hash-filenames
262 By default, gcov uses the full pathname of the source files to to
263 create an output filename. This can lead to long filenames that
264 can overflow filesystem limits. This option creates names of the
265 form source-file##md5.gcov, where the source-file component is the
266 final filename part and the md5 component is calculated from the
267 full mangled name that would have been used otherwise.
268 gcov should be run with the current directory the same as that when you
270 invoked the compiler. Otherwise it will not be able to locate the
271 source files. gcov produces files called mangledname.gcov in the
272 current directory. These contain the coverage information of the
273 source file they correspond to. One .gcov file is produced for each
274 source (or header) file containing code, which was compiled to produce
275 the data files. The mangledname part of the output file name is
276 usually simply the source file name, but can be something more
277 complicated if the -l or -p options are given. Refer to those options
278 for details.
279 If you invoke gcov with multiple input files, the contributions from
281 each input file are summed. Typically you would invoke it with the
282 same list of files as the final link of your executable.
283 The .gcov files contain the : separated fields along with program
285 source code. The format is
286 <execution_count>:<line_number>:<source line text>
288 Additional block information may succeed each line, when requested by
290 command line option. The execution_count is - for lines containing no
291 code. Unexecuted lines are marked ##### or =====, depending on whether
292 they are reachable by non-exceptional paths or only exceptional paths
293 such as C++ exception handlers, respectively. Given -a option,
294 unexecuted blocks are marked $$$$$ or %%%%%, depending on whether a
295 basic block is reachable via non-exceptional or exceptional paths.
296 Executed basic blocks having a statement with zero execution_count end
297 with * character and are colored with magenta color with -k option.
298 The functionality is not supported in Ada.
299 Note that GCC can completely remove the bodies of functions that are
301 not needed -- for instance if they are inlined everywhere. Such
302 functions are marked with -, which can be confusing. Use the
303 -fkeep-inline-functions and -fkeep-static-functions options to retain
304 these functions and allow gcov to properly show their execution_count.
305 Some lines of information at the start have line_number of zero. These
307 preamble lines are of the form
308 -:0:<tag>:<value>
310 The ordering and number of these preamble lines will be augmented as
312 gcov development progresses --- do not rely on them remaining
313 unchanged. Use tag to locate a particular preamble line.
314 The additional block information is of the form
316 <tag> <information>
318 The information is human readable, but designed to be simple enough for
320 machine parsing too.
321 When printing percentages, 0% and 100% are only printed when the values
323 are exactly 0% and 100% respectively. Other values which would
324 conventionally be rounded to 0% or 100% are instead printed as the
325 nearest non-boundary value.
326 When using gcov, you must first compile your program with two special
328 GCC options: -fprofile-arcs -ftest-coverage. This tells the compiler
329 to generate additional information needed by gcov (basically a flow
330 graph of the program) and also includes additional code in the object
331 files for generating the extra profiling information needed by gcov.
332 These additional files are placed in the directory where the object
333 file is located.
334 Running the program will cause profile output to be generated. For
336 each source file compiled with -fprofile-arcs, an accompanying .gcda
337 file will be placed in the object file directory.
338 Running gcov with your program's source file names as arguments will
340 now produce a listing of the code along with frequency of execution for
341 each line. For example, if your program is called tmp.cpp, this is
342 what you see when you use the basic gcov facility:
343 $ g++ -fprofile-arcs -ftest-coverage tmp.cpp
345 $ a.out
346 $ gcov tmp.cpp -m
347 File 'tmp.cpp'
348 Lines executed:92.86% of 14
349 Creating 'tmp.cpp.gcov'
350 The file tmp.cpp.gcov contains output from gcov. Here is a sample:
352 -: 0:Source:tmp.cpp
354 -: 0:Graph:tmp.gcno
355 -: 0:Data:tmp.gcda
356 -: 0:Runs:1
357 -: 0:Programs:1
358 -: 1:#include <stdio.h>
359 -: 2:
360 -: 3:template<class T>
361 -: 4:class Foo
362 -: 5:{
363 -: 6: public:
364 1*: 7: Foo(): b (1000) {}
365 ------------------
366 Foo<char>::Foo():
367 #####: 7: Foo(): b (1000) {}
368 ------------------
369 Foo<int>::Foo():
370 1: 7: Foo(): b (1000) {}
371 ------------------
372 2*: 8: void inc () { b++; }
373 ------------------
374 Foo<char>::inc():
375 #####: 8: void inc () { b++; }
376 ------------------
377 Foo<int>::inc():
378 2: 8: void inc () { b++; }
379 ------------------
380 -: 9:
381 -: 10: private:
382 -: 11: int b;
383 -: 12:};
384 -: 13:
385 -: 14:template class Foo<int>;
386 -: 15:template class Foo<char>;
387 -: 16:
388 -: 17:int
389 1: 18:main (void)
390 -: 19:{
391 -: 20: int i, total;
392 1: 21: Foo<int> counter;
393 -: 22:
394 1: 23: counter.inc();
395 1: 24: counter.inc();
396 1: 25: total = 0;
397 -: 26:
398 11: 27: for (i = 0; i < 10; i++)
399 10: 28: total += i;
400 -: 29:
401 1*: 30: int v = total > 100 ? 1 : 2;
402 -: 31:
403 1: 32: if (total != 45)
404 #####: 33: printf ("Failure\n");
405 -: 34: else
406 1: 35: printf ("Success\n");
407 1: 36: return 0;
408 -: 37:}
409 Note that line 7 is shown in the report multiple times. First
411 occurrence presents total number of execution of the line and the next
412 two belong to instances of class Foo constructors. As you can also
413 see, line 30 contains some unexecuted basic blocks and thus execution
414 count has asterisk symbol.
415 When you use the -a option, you will get individual block counts, and
417 the output looks like this:
418 -: 0:Source:tmp.cpp
420 -: 0:Graph:tmp.gcno
421 -: 0:Data:tmp.gcda
422 -: 0:Runs:1
423 -: 0:Programs:1
424 -: 1:#include <stdio.h>
425 -: 2:
426 -: 3:template<class T>
427 -: 4:class Foo
428 -: 5:{
429 -: 6: public:
430 1*: 7: Foo(): b (1000) {}
431 ------------------
432 Foo<char>::Foo():
433 #####: 7: Foo(): b (1000) {}
434 ------------------
435 Foo<int>::Foo():
436 1: 7: Foo(): b (1000) {}
437 ------------------
438 2*: 8: void inc () { b++; }
439 ------------------
440 Foo<char>::inc():
441 #####: 8: void inc () { b++; }
442 ------------------
443 Foo<int>::inc():
444 2: 8: void inc () { b++; }
445 ------------------
446 -: 9:
447 -: 10: private:
448 -: 11: int b;
449 -: 12:};
450 -: 13:
451 -: 14:template class Foo<int>;
452 -: 15:template class Foo<char>;
453 -: 16:
454 -: 17:int
455 1: 18:main (void)
456 -: 19:{
457 -: 20: int i, total;
458 1: 21: Foo<int> counter;
459 1: 21-block 0
460 -: 22:
461 1: 23: counter.inc();
462 1: 23-block 0
463 1: 24: counter.inc();
464 1: 24-block 0
465 1: 25: total = 0;
466 -: 26:
467 11: 27: for (i = 0; i < 10; i++)
468 1: 27-block 0
469 11: 27-block 1
470 10: 28: total += i;
471 10: 28-block 0
472 -: 29:
473 1*: 30: int v = total > 100 ? 1 : 2;
474 1: 30-block 0
475 %%%%%: 30-block 1
476 1: 30-block 2
477 -: 31:
478 1: 32: if (total != 45)
479 1: 32-block 0
480 #####: 33: printf ("Failure\n");
481 %%%%%: 33-block 0
482 -: 34: else
483 1: 35: printf ("Success\n");
484 1: 35-block 0
485 1: 36: return 0;
486 1: 36-block 0
487 -: 37:}
488 In this mode, each basic block is only shown on one line -- the last
490 line of the block. A multi-line block will only contribute to the
491 execution count of that last line, and other lines will not be shown to
492 contain code, unless previous blocks end on those lines. The total
493 execution count of a line is shown and subsequent lines show the
494 execution counts for individual blocks that end on that line. After
495 each block, the branch and call counts of the block will be shown, if
496 the -b option is given.
497 Because of the way GCC instruments calls, a call count can be shown
499 after a line with no individual blocks. As you can see, line 33
500 contains a basic block that was not executed.
501 When you use the -b option, your output looks like this:
503 -: 0:Source:tmp.cpp
505 -: 0:Graph:tmp.gcno
506 -: 0:Data:tmp.gcda
507 -: 0:Runs:1
508 -: 0:Programs:1
509 -: 1:#include <stdio.h>
510 -: 2:
511 -: 3:template<class T>
512 -: 4:class Foo
513 -: 5:{
514 -: 6: public:
515 1*: 7: Foo(): b (1000) {}
516 ------------------
517 Foo<char>::Foo():
518 function Foo<char>::Foo() called 0 returned 0% blocks executed 0%
519 #####: 7: Foo(): b (1000) {}
520 ------------------
521 Foo<int>::Foo():
522 function Foo<int>::Foo() called 1 returned 100% blocks executed 100%
523 1: 7: Foo(): b (1000) {}
524 ------------------
525 2*: 8: void inc () { b++; }
526 ------------------
527 Foo<char>::inc():
528 function Foo<char>::inc() called 0 returned 0% blocks executed 0%
529 #####: 8: void inc () { b++; }
530 ------------------
531 Foo<int>::inc():
532 function Foo<int>::inc() called 2 returned 100% blocks executed 100%
533 2: 8: void inc () { b++; }
534 ------------------
535 -: 9:
536 -: 10: private:
537 -: 11: int b;
538 -: 12:};
539 -: 13:
540 -: 14:template class Foo<int>;
541 -: 15:template class Foo<char>;
542 -: 16:
543 -: 17:int
544 function main called 1 returned 100% blocks executed 81%
545 1: 18:main (void)
546 -: 19:{
547 -: 20: int i, total;
548 1: 21: Foo<int> counter;
549 call 0 returned 100%
550 branch 1 taken 100% (fallthrough)
551 branch 2 taken 0% (throw)
552 -: 22:
553 1: 23: counter.inc();
554 call 0 returned 100%
555 branch 1 taken 100% (fallthrough)
556 branch 2 taken 0% (throw)
557 1: 24: counter.inc();
558 call 0 returned 100%
559 branch 1 taken 100% (fallthrough)
560 branch 2 taken 0% (throw)
561 1: 25: total = 0;
562 -: 26:
563 11: 27: for (i = 0; i < 10; i++)
564 branch 0 taken 91% (fallthrough)
565 branch 1 taken 9%
566 10: 28: total += i;
567 -: 29:
568 1*: 30: int v = total > 100 ? 1 : 2;
569 branch 0 taken 0% (fallthrough)
570 branch 1 taken 100%
571 -: 31:
572 1: 32: if (total != 45)
573 branch 0 taken 0% (fallthrough)
574 branch 1 taken 100%
575 #####: 33: printf ("Failure\n");
576 call 0 never executed
577 branch 1 never executed
578 branch 2 never executed
579 -: 34: else
580 1: 35: printf ("Success\n");
581 call 0 returned 100%
582 branch 1 taken 100% (fallthrough)
583 branch 2 taken 0% (throw)
584 1: 36: return 0;
585 -: 37:}
586 For each function, a line is printed showing how many times the
588 function is called, how many times it returns and what percentage of
589 the function's blocks were executed.
590 For each basic block, a line is printed after the last line of the
592 basic block describing the branch or call that ends the basic block.
593 There can be multiple branches and calls listed for a single source
594 line if there are multiple basic blocks that end on that line. In this
595 case, the branches and calls are each given a number. There is no
596 simple way to map these branches and calls back to source constructs.
597 In general, though, the lowest numbered branch or call will correspond
598 to the leftmost construct on the source line.
599 For a branch, if it was executed at least once, then a percentage
601 indicating the number of times the branch was taken divided by the
602 number of times the branch was executed will be printed. Otherwise,
603 the message "never executed" is printed.
604 For a call, if it was executed at least once, then a percentage
606 indicating the number of times the call returned divided by the number
607 of times the call was executed will be printed. This will usually be
608 100%, but may be less for functions that call "exit" or "longjmp", and
609 thus may not return every time they are called.
610 The execution counts are cumulative. If the example program were
612 executed again without removing the .gcda file, the count for the
613 number of times each line in the source was executed would be added to
614 the results of the previous run(s). This is potentially useful in
615 several ways. For example, it could be used to accumulate data over a
616 number of program runs as part of a test verification suite, or to
617 provide more accurate long-term information over a large number of
618 program runs.
619 The data in the .gcda files is saved immediately before the program
621 exits. For each source file compiled with -fprofile-arcs, the
622 profiling code first attempts to read in an existing .gcda file; if the
623 file doesn't match the executable (differing number of basic block
624 counts) it will ignore the contents of the file. It then adds in the
625 new execution counts and finally writes the data to the file.
626 Using gcov with GCC Optimization
628 If you plan to use gcov to help optimize your code, you must first
629 compile your program with two special GCC options: -fprofile-arcs
630 -ftest-coverage. Aside from that, you can use any other GCC options;
631 but if you want to prove that every single line in your program was
632 executed, you should not compile with optimization at the same time.
633 On some machines the optimizer can eliminate some simple code lines by
634 combining them with other lines. For example, code like this:
635 if (a != b)
637 c = 1;
638 else
639 c = 0;
640 can be compiled into one instruction on some machines. In this case,
642 there is no way for gcov to calculate separate execution counts for
643 each line because there isn't separate code for each line. Hence the
644 gcov output looks like this if you compiled the program with
645 optimization:
646 100: 12:if (a != b)
648 100: 13: c = 1;
649 100: 14:else
650 100: 15: c = 0;
651 The output shows that this block of code, combined by optimization,
653 executed 100 times. In one sense this result is correct, because there
654 was only one instruction representing all four of these lines.
655 However, the output does not indicate how many times the result was 0
656 and how many times the result was 1.
657 Inlineable functions can create unexpected line counts. Line counts
659 are shown for the source code of the inlineable function, but what is
660 shown depends on where the function is inlined, or if it is not inlined
661 at all.
662 If the function is not inlined, the compiler must emit an out of line
664 copy of the function, in any object file that needs it. If fileA.o and
665 fileB.o both contain out of line bodies of a particular inlineable
666 function, they will also both contain coverage counts for that
667 function. When fileA.o and fileB.o are linked together, the linker
668 will, on many systems, select one of those out of line bodies for all
669 calls to that function, and remove or ignore the other. Unfortunately,
670 it will not remove the coverage counters for the unused function body.
671 Hence when instrumented, all but one use of that function will show
672 zero counts.
673 If the function is inlined in several places, the block structure in
675 each location might not be the same. For instance, a condition might
676 now be calculable at compile time in some instances. Because the
677 coverage of all the uses of the inline function will be shown for the
678 same source lines, the line counts themselves might seem inconsistent.
679 Long-running applications can use the "__gcov_reset" and "__gcov_dump"
681 facilities to restrict profile collection to the program region of
682 interest. Calling "__gcov_reset(void)" will clear all profile counters
683 to zero, and calling "__gcov_dump(void)" will cause the profile
684 information collected at that point to be dumped to .gcda output files.
685 Instrumented applications use a static destructor with priority 99 to
686 invoke the "__gcov_dump" function. Thus "__gcov_dump" is executed after
687 all user defined static destructors, as well as handlers registered
688 with "atexit". If an executable loads a dynamic shared object via
689 dlopen functionality, -Wl,--dynamic-list-data is needed to dump all
690 profile data.
691
693 gpl(7), gfdl(7), fsf-funding(7), gcc(1) and the Info entry for gcc.
694
696 Copyright (c) 1996-2018 Free Software Foundation, Inc.
697 Permission is granted to copy, distribute and/or modify this document
699 under the terms of the GNU Free Documentation License, Version 1.3 or
700 any later version published by the Free Software Foundation; with the
701 Invariant Sections being "GNU General Public License" and "Funding Free
702 Software", the Front-Cover texts being (a) (see below), and with the
703 Back-Cover Texts being (b) (see below). A copy of the license is
704 included in the gfdl(7) man page.
705 (a) The FSF's Front-Cover Text is:
707 A GNU Manual
709 (b) The FSF's Back-Cover Text is:
711 You have freedom to copy and modify this GNU Manual, like GNU
713 software. Copies published by the Free Software Foundation raise
714 funds for GNU development.
715gcc-8 2018-06-26 GCOV(1)