1@c Copyright (C) 1996, 1997, 1999, 2000, 2001, 2@c 2002, 2003, 2004, 2005 Free Software Foundation, Inc. 3@c This is part of the GCC manual. 4@c For copying conditions, see the file gcc.texi. 5 6@ignore 7@c man begin COPYRIGHT 8Copyright @copyright{} 1996, 1997, 1999, 2000, 2001, 2002, 2003, 2004, 2005 9Free Software Foundation, Inc. 10 11Permission is granted to copy, distribute and/or modify this document 12under the terms of the GNU Free Documentation License, Version 1.2 or 13any later version published by the Free Software Foundation; with the 14Invariant Sections being ``GNU General Public License'' and ``Funding 15Free Software'', the Front-Cover texts being (a) (see below), and with 16the Back-Cover Texts being (b) (see below). A copy of the license is 17included in the gfdl(7) man page. 18 19(a) The FSF's Front-Cover Text is: 20 21 A GNU Manual 22 23(b) The FSF's Back-Cover Text is: 24 25 You have freedom to copy and modify this GNU Manual, like GNU 26 software. Copies published by the Free Software Foundation raise 27 funds for GNU development. 28@c man end 29@c Set file name and title for the man page. 30@setfilename gcov 31@settitle coverage testing tool 32@end ignore 33 34@node Gcov 35@chapter @command{gcov}---a Test Coverage Program 36 37@command{gcov} is a tool you can use in conjunction with GCC to 38test code coverage in your programs. 39 40@menu 41* Gcov Intro:: Introduction to gcov. 42* Invoking Gcov:: How to use gcov. 43* Gcov and Optimization:: Using gcov with GCC optimization. 44* Gcov Data Files:: The files used by gcov. 45* Cross-profiling:: Data file relocation. 46@end menu 47 48@node Gcov Intro 49@section Introduction to @command{gcov} 50@c man begin DESCRIPTION 51 52@command{gcov} is a test coverage program. Use it in concert with GCC 53to analyze your programs to help create more efficient, faster running 54code and to discover untested parts of your program. You can use 55@command{gcov} as a profiling tool to help discover where your 56optimization efforts will best affect your code. You can also use 57@command{gcov} along with the other profiling tool, @command{gprof}, to 58assess which parts of your code use the greatest amount of computing 59time. 60 61Profiling tools help you analyze your code's performance. Using a 62profiler such as @command{gcov} or @command{gprof}, you can find out some 63basic performance statistics, such as: 64 65@itemize @bullet 66@item 67how often each line of code executes 68 69@item 70what lines of code are actually executed 71 72@item 73how much computing time each section of code uses 74@end itemize 75 76Once you know these things about how your code works when compiled, you 77can look at each module to see which modules should be optimized. 78@command{gcov} helps you determine where to work on optimization. 79 80Software developers also use coverage testing in concert with 81testsuites, to make sure software is actually good enough for a release. 82Testsuites can verify that a program works as expected; a coverage 83program tests to see how much of the program is exercised by the 84testsuite. Developers can then determine what kinds of test cases need 85to be added to the testsuites to create both better testing and a better 86final product. 87 88You should compile your code without optimization if you plan to use 89@command{gcov} because the optimization, by combining some lines of code 90into one function, may not give you as much information as you need to 91look for `hot spots' where the code is using a great deal of computer 92time. Likewise, because @command{gcov} accumulates statistics by line (at 93the lowest resolution), it works best with a programming style that 94places only one statement on each line. If you use complicated macros 95that expand to loops or to other control structures, the statistics are 96less helpful---they only report on the line where the macro call 97appears. If your complex macros behave like functions, you can replace 98them with inline functions to solve this problem. 99 100@command{gcov} creates a logfile called @file{@var{sourcefile}.gcov} which 101indicates how many times each line of a source file @file{@var{sourcefile}.c} 102has executed. You can use these logfiles along with @command{gprof} to aid 103in fine-tuning the performance of your programs. @command{gprof} gives 104timing information you can use along with the information you get from 105@command{gcov}. 106 107@command{gcov} works only on code compiled with GCC@. It is not 108compatible with any other profiling or test coverage mechanism. 109 110@c man end 111 112@node Invoking Gcov 113@section Invoking @command{gcov} 114 115@smallexample 116gcov @r{[}@var{options}@r{]} @var{sourcefile} 117@end smallexample 118 119@command{gcov} accepts the following options: 120 121@ignore 122@c man begin SYNOPSIS 123gcov [@option{-v}|@option{--version}] [@option{-h}|@option{--help}] 124 [@option{-a}|@option{--all-blocks}] 125 [@option{-b}|@option{--branch-probabilities}] 126 [@option{-c}|@option{--branch-counts}] 127 [@option{-n}|@option{--no-output}] 128 [@option{-l}|@option{--long-file-names}] 129 [@option{-p}|@option{--preserve-paths}] 130 [@option{-f}|@option{--function-summaries}] 131 [@option{-o}|@option{--object-directory} @var{directory|file}] @var{sourcefile} 132 [@option{-u}|@option{--unconditional-branches}] 133@c man end 134@c man begin SEEALSO 135gpl(7), gfdl(7), fsf-funding(7), gcc(1) and the Info entry for @file{gcc}. 136@c man end 137@end ignore 138 139@c man begin OPTIONS 140@table @gcctabopt 141@item -h 142@itemx --help 143Display help about using @command{gcov} (on the standard output), and 144exit without doing any further processing. 145 146@item -v 147@itemx --version 148Display the @command{gcov} version number (on the standard output), 149and exit without doing any further processing. 150 151@item -a 152@itemx --all-blocks 153Write individual execution counts for every basic block. Normally gcov 154outputs execution counts only for the main blocks of a line. With this 155option you can determine if blocks within a single line are not being 156executed. 157 158@item -b 159@itemx --branch-probabilities 160Write branch frequencies to the output file, and write branch summary 161info to the standard output. This option allows you to see how often 162each branch in your program was taken. Unconditional branches will not 163be shown, unless the @option{-u} option is given. 164 165@item -c 166@itemx --branch-counts 167Write branch frequencies as the number of branches taken, rather than 168the percentage of branches taken. 169 170@item -n 171@itemx --no-output 172Do not create the @command{gcov} output file. 173 174@item -l 175@itemx --long-file-names 176Create long file names for included source files. For example, if the 177header file @file{x.h} contains code, and was included in the file 178@file{a.c}, then running @command{gcov} on the file @file{a.c} will produce 179an output file called @file{a.c##x.h.gcov} instead of @file{x.h.gcov}. 180This can be useful if @file{x.h} is included in multiple source 181files. If you use the @samp{-p} option, both the including and 182included file names will be complete path names. 183 184@item -p 185@itemx --preserve-paths 186Preserve complete path information in the names of generated 187@file{.gcov} files. Without this option, just the filename component is 188used. With this option, all directories are used, with @samp{/} characters 189translated to @samp{#} characters, @file{.} directory components 190removed and @file{..} 191components renamed to @samp{^}. This is useful if sourcefiles are in several 192different directories. It also affects the @samp{-l} option. 193 194@item -f 195@itemx --function-summaries 196Output summaries for each function in addition to the file level summary. 197 198@item -o @var{directory|file} 199@itemx --object-directory @var{directory} 200@itemx --object-file @var{file} 201Specify either the directory containing the gcov data files, or the 202object path name. The @file{.gcno}, and 203@file{.gcda} data files are searched for using this option. If a directory 204is specified, the data files are in that directory and named after the 205source file name, without its extension. If a file is specified here, 206the data files are named after that file, without its extension. If this 207option is not supplied, it defaults to the current directory. 208 209@item -u 210@itemx --unconditional-branches 211When branch probabilities are given, include those of unconditional branches. 212Unconditional branches are normally not interesting. 213 214@end table 215 216@command{gcov} should be run with the current directory the same as that 217when you invoked the compiler. Otherwise it will not be able to locate 218the source files. @command{gcov} produces files called 219@file{@var{mangledname}.gcov} in the current directory. These contain 220the coverage information of the source file they correspond to. 221One @file{.gcov} file is produced for each source file containing code, 222which was compiled to produce the data files. The @var{mangledname} part 223of the output file name is usually simply the source file name, but can 224be something more complicated if the @samp{-l} or @samp{-p} options are 225given. Refer to those options for details. 226 227The @file{.gcov} files contain the @samp{:} separated fields along with 228program source code. The format is 229 230@smallexample 231@var{execution_count}:@var{line_number}:@var{source line text} 232@end smallexample 233 234Additional block information may succeed each line, when requested by 235command line option. The @var{execution_count} is @samp{-} for lines 236containing no code and @samp{#####} for lines which were never executed. 237Some lines of information at the start have @var{line_number} of zero. 238 239The preamble lines are of the form 240 241@smallexample 242-:0:@var{tag}:@var{value} 243@end smallexample 244 245The ordering and number of these preamble lines will be augmented as 246@command{gcov} development progresses --- do not rely on them remaining 247unchanged. Use @var{tag} to locate a particular preamble line. 248 249The additional block information is of the form 250 251@smallexample 252@var{tag} @var{information} 253@end smallexample 254 255The @var{information} is human readable, but designed to be simple 256enough for machine parsing too. 257 258When printing percentages, 0% and 100% are only printed when the values 259are @emph{exactly} 0% and 100% respectively. Other values which would 260conventionally be rounded to 0% or 100% are instead printed as the 261nearest non-boundary value. 262 263When using @command{gcov}, you must first compile your program with two 264special GCC options: @samp{-fprofile-arcs -ftest-coverage}. 265This tells the compiler to generate additional information needed by 266gcov (basically a flow graph of the program) and also includes 267additional code in the object files for generating the extra profiling 268information needed by gcov. These additional files are placed in the 269directory where the object file is located. 270 271Running the program will cause profile output to be generated. For each 272source file compiled with @option{-fprofile-arcs}, an accompanying 273@file{.gcda} file will be placed in the object file directory. 274 275Running @command{gcov} with your program's source file names as arguments 276will now produce a listing of the code along with frequency of execution 277for each line. For example, if your program is called @file{tmp.c}, this 278is what you see when you use the basic @command{gcov} facility: 279 280@smallexample 281$ gcc -fprofile-arcs -ftest-coverage tmp.c 282$ a.out 283$ gcov tmp.c 28490.00% of 10 source lines executed in file tmp.c 285Creating tmp.c.gcov. 286@end smallexample 287 288The file @file{tmp.c.gcov} contains output from @command{gcov}. 289Here is a sample: 290 291@smallexample 292 -: 0:Source:tmp.c 293 -: 0:Graph:tmp.gcno 294 -: 0:Data:tmp.gcda 295 -: 0:Runs:1 296 -: 0:Programs:1 297 -: 1:#include <stdio.h> 298 -: 2: 299 -: 3:int main (void) 300 1: 4:@{ 301 1: 5: int i, total; 302 -: 6: 303 1: 7: total = 0; 304 -: 8: 305 11: 9: for (i = 0; i < 10; i++) 306 10: 10: total += i; 307 -: 11: 308 1: 12: if (total != 45) 309 #####: 13: printf ("Failure\n"); 310 -: 14: else 311 1: 15: printf ("Success\n"); 312 1: 16: return 0; 313 -: 17:@} 314@end smallexample 315 316When you use the @option{-a} option, you will get individual block 317counts, and the output looks like this: 318 319@smallexample 320 -: 0:Source:tmp.c 321 -: 0:Graph:tmp.gcno 322 -: 0:Data:tmp.gcda 323 -: 0:Runs:1 324 -: 0:Programs:1 325 -: 1:#include <stdio.h> 326 -: 2: 327 -: 3:int main (void) 328 1: 4:@{ 329 1: 4-block 0 330 1: 5: int i, total; 331 -: 6: 332 1: 7: total = 0; 333 -: 8: 334 11: 9: for (i = 0; i < 10; i++) 335 11: 9-block 0 336 10: 10: total += i; 337 10: 10-block 0 338 -: 11: 339 1: 12: if (total != 45) 340 1: 12-block 0 341 #####: 13: printf ("Failure\n"); 342 $$$$$: 13-block 0 343 -: 14: else 344 1: 15: printf ("Success\n"); 345 1: 15-block 0 346 1: 16: return 0; 347 1: 16-block 0 348 -: 17:@} 349@end smallexample 350 351In this mode, each basic block is only shown on one line -- the last 352line of the block. A multi-line block will only contribute to the 353execution count of that last line, and other lines will not be shown 354to contain code, unless previous blocks end on those lines. 355The total execution count of a line is shown and subsequent lines show 356the execution counts for individual blocks that end on that line. After each 357block, the branch and call counts of the block will be shown, if the 358@option{-b} option is given. 359 360Because of the way GCC instruments calls, a call count can be shown 361after a line with no individual blocks. 362As you can see, line 13 contains a basic block that was not executed. 363 364@need 450 365When you use the @option{-b} option, your output looks like this: 366 367@smallexample 368$ gcov -b tmp.c 36990.00% of 10 source lines executed in file tmp.c 37080.00% of 5 branches executed in file tmp.c 37180.00% of 5 branches taken at least once in file tmp.c 37250.00% of 2 calls executed in file tmp.c 373Creating tmp.c.gcov. 374@end smallexample 375 376Here is a sample of a resulting @file{tmp.c.gcov} file: 377 378@smallexample 379 -: 0:Source:tmp.c 380 -: 0:Graph:tmp.gcno 381 -: 0:Data:tmp.gcda 382 -: 0:Runs:1 383 -: 0:Programs:1 384 -: 1:#include <stdio.h> 385 -: 2: 386 -: 3:int main (void) 387function main called 1 returned 1 blocks executed 75% 388 1: 4:@{ 389 1: 5: int i, total; 390 -: 6: 391 1: 7: total = 0; 392 -: 8: 393 11: 9: for (i = 0; i < 10; i++) 394branch 0 taken 91% (fallthrough) 395branch 1 taken 9% 396 10: 10: total += i; 397 -: 11: 398 1: 12: if (total != 45) 399branch 0 taken 0% (fallthrough) 400branch 1 taken 100% 401 #####: 13: printf ("Failure\n"); 402call 0 never executed 403 -: 14: else 404 1: 15: printf ("Success\n"); 405call 0 called 1 returned 100% 406 1: 16: return 0; 407 -: 17:@} 408@end smallexample 409 410For each function, a line is printed showing how many times the function 411is called, how many times it returns and what percentage of the 412function's blocks were executed. 413 414For each basic block, a line is printed after the last line of the basic 415block describing the branch or call that ends the basic block. There can 416be multiple branches and calls listed for a single source line if there 417are multiple basic blocks that end on that line. In this case, the 418branches and calls are each given a number. There is no simple way to map 419these branches and calls back to source constructs. In general, though, 420the lowest numbered branch or call will correspond to the leftmost construct 421on the source line. 422 423For a branch, if it was executed at least once, then a percentage 424indicating the number of times the branch was taken divided by the 425number of times the branch was executed will be printed. Otherwise, the 426message ``never executed'' is printed. 427 428For a call, if it was executed at least once, then a percentage 429indicating the number of times the call returned divided by the number 430of times the call was executed will be printed. This will usually be 431100%, but may be less for functions that call @code{exit} or @code{longjmp}, 432and thus may not return every time they are called. 433 434The execution counts are cumulative. If the example program were 435executed again without removing the @file{.gcda} file, the count for the 436number of times each line in the source was executed would be added to 437the results of the previous run(s). This is potentially useful in 438several ways. For example, it could be used to accumulate data over a 439number of program runs as part of a test verification suite, or to 440provide more accurate long-term information over a large number of 441program runs. 442 443The data in the @file{.gcda} files is saved immediately before the program 444exits. For each source file compiled with @option{-fprofile-arcs}, the 445profiling code first attempts to read in an existing @file{.gcda} file; if 446the file doesn't match the executable (differing number of basic block 447counts) it will ignore the contents of the file. It then adds in the 448new execution counts and finally writes the data to the file. 449 450@node Gcov and Optimization 451@section Using @command{gcov} with GCC Optimization 452 453If you plan to use @command{gcov} to help optimize your code, you must 454first compile your program with two special GCC options: 455@samp{-fprofile-arcs -ftest-coverage}. Aside from that, you can use any 456other GCC options; but if you want to prove that every single line 457in your program was executed, you should not compile with optimization 458at the same time. On some machines the optimizer can eliminate some 459simple code lines by combining them with other lines. For example, code 460like this: 461 462@smallexample 463if (a != b) 464 c = 1; 465else 466 c = 0; 467@end smallexample 468 469@noindent 470can be compiled into one instruction on some machines. In this case, 471there is no way for @command{gcov} to calculate separate execution counts 472for each line because there isn't separate code for each line. Hence 473the @command{gcov} output looks like this if you compiled the program with 474optimization: 475 476@smallexample 477 100: 12:if (a != b) 478 100: 13: c = 1; 479 100: 14:else 480 100: 15: c = 0; 481@end smallexample 482 483The output shows that this block of code, combined by optimization, 484executed 100 times. In one sense this result is correct, because there 485was only one instruction representing all four of these lines. However, 486the output does not indicate how many times the result was 0 and how 487many times the result was 1. 488 489Inlineable functions can create unexpected line counts. Line counts are 490shown for the source code of the inlineable function, but what is shown 491depends on where the function is inlined, or if it is not inlined at all. 492 493If the function is not inlined, the compiler must emit an out of line 494copy of the function, in any object file that needs it. If 495@file{fileA.o} and @file{fileB.o} both contain out of line bodies of a 496particular inlineable function, they will also both contain coverage 497counts for that function. When @file{fileA.o} and @file{fileB.o} are 498linked together, the linker will, on many systems, select one of those 499out of line bodies for all calls to that function, and remove or ignore 500the other. Unfortunately, it will not remove the coverage counters for 501the unused function body. Hence when instrumented, all but one use of 502that function will show zero counts. 503 504If the function is inlined in several places, the block structure in 505each location might not be the same. For instance, a condition might 506now be calculable at compile time in some instances. Because the 507coverage of all the uses of the inline function will be shown for the 508same source lines, the line counts themselves might seem inconsistent. 509 510@c man end 511 512@node Gcov Data Files 513@section Brief description of @command{gcov} data files 514 515@command{gcov} uses two files for profiling. The names of these files 516are derived from the original @emph{object} file by substituting the 517file suffix with either @file{.gcno}, or @file{.gcda}. All of these files 518are placed in the same directory as the object file, and contain data 519stored in a platform-independent format. 520 521The @file{.gcno} file is generated when the source file is compiled with 522the GCC @option{-ftest-coverage} option. It contains information to 523reconstruct the basic block graphs and assign source line numbers to 524blocks. 525 526The @file{.gcda} file is generated when a program containing object files 527built with the GCC @option{-fprofile-arcs} option is executed. A 528separate @file{.gcda} file is created for each object file compiled with 529this option. It contains arc transition counts, and some summary 530information. 531 532The full details of the file format is specified in @file{gcov-io.h}, 533and functions provided in that header file should be used to access the 534coverage files. 535 536@node Cross-profiling 537@section Data file relocation to support cross-profiling 538 539Running the program will cause profile output to be generated. For each 540source file compiled with @option{-fprofile-arcs}, an accompanying @file{.gcda} 541file will be placed in the object file directory. That implicitly requires 542running the program on the same system as it was built or having the same 543absolute directory structure on the target system. The program will try 544to create the needed directory structure, if it is not already present. 545 546To support cross-profiling, a program compiled with @option{-fprofile-arcs} 547can relocate the data files based on two environment variables: 548 549@itemize @bullet 550@item 551GCOV_PREFIX contains the prefix to add to the absolute paths 552in the object file. Prefix must be absolute as well, otherwise its 553value is ignored. The default is no prefix. 554 555@item 556GCOV_PREFIX_STRIP indicates the how many initial directory names to strip off 557the hardwired absolute paths. Default value is 0. 558 559@emph{Note:} GCOV_PREFIX_STRIP has no effect if GCOV_PREFIX is undefined, empty 560or non-absolute. 561@end itemize 562 563For example, if the object file @file{/user/build/foo.o} was built with 564@option{-fprofile-arcs}, the final executable will try to create the data file 565@file{/user/build/foo.gcda} when running on the target system. This will 566fail if the corresponding directory does not exist and it is unable to create 567it. This can be overcome by, for example, setting the environment as 568@samp{GCOV_PREFIX=/target/run} and @samp{GCOV_PREFIX_STRIP=1}. Such a 569setting will name the data file @file{/target/run/build/foo.gcda}. 570 571You must move the data files to the expected directory tree in order to 572use them for profile directed optimizations (@option{--use-profile}), or to 573use the @command{gcov} tool. 574