1LLVM-COV(1)                          LLVM                          LLVM-COV(1)
2
3
4

NAME

6       llvm-cov - emit coverage information
7

SYNOPSIS

9       llvm-cov command [args...]
10

DESCRIPTION

12       The llvm-cov tool shows code coverage information for programs that are
13       instrumented to emit  profile  data.  It  can  be  used  to  work  with
14       gcov-style coverage or with clang's instrumentation based profiling.
15
16       If  the  program is invoked with a base name of gcov, it will behave as
17       if the llvm-cov gcov command were called. Otherwise, a  command  should
18       be provided.
19

COMMANDS

21gcov
22
23show
24
25report
26
27export
28

GCOV COMMAND

30   SYNOPSIS
31       llvm-cov gcov [options] SOURCEFILE
32
33   DESCRIPTION
34       The  llvm-cov gcov tool reads code coverage data files and displays the
35       coverage information for a specified source file. It is compatible with
36       the  gcov  tool from version 4.2 of GCC and may also be compatible with
37       some later versions of gcov.
38
39       To use llvm-cov gcov, you must first build an instrumented  version  of
40       your  application  that collects coverage data as it runs. Compile with
41       the -fprofile-arcs and -ftest-coverage options to add the  instrumenta‐
42       tion. (Alternatively, you can use the --coverage option, which includes
43       both of those other options.) You should compile with debugging  infor‐
44       mation  (-g)  and  without  optimization (-O0); otherwise, the coverage
45       data cannot be accurately mapped back to the source code.
46
47       At the time you compile the instrumented code, a .gcno data  file  will
48       be  generated  for  each object file. These .gcno files contain half of
49       the coverage data. The other half of the data comes  from  .gcda  files
50       that  are generated when you run the instrumented program, with a sepa‐
51       rate .gcda file for each object file. Each time you  run  the  program,
52       the  execution  counts  are summed into any existing .gcda files, so be
53       sure to remove any old files if you do not want their  contents  to  be
54       included.
55
56       By  default, the .gcda files are written into the same directory as the
57       object files, but you can override that by setting the GCOV_PREFIX  and
58       GCOV_PREFIX_STRIP environment variables. The GCOV_PREFIX_STRIP variable
59       specifies a number of directory components to be removed from the start
60       of  the  absolute  path  to  the object file directory. After stripping
61       those directories, the prefix from the GCOV_PREFIX variable  is  added.
62       These  environment  variables allow you to run the instrumented program
63       on a machine where the original object file directories are not  acces‐
64       sible,  but  you will then need to copy the .gcda files back to the ob‐
65       ject file directories where llvm-cov gcov expects to find them.
66
67       Once you have generated the coverage data files, run llvm-cov gcov  for
68       each  main  source file where you want to examine the coverage results.
69       This should be run from the same directory where you previously ran the
70       compiler.  The  results  for the specified source file are written to a
71       file named by appending a .gcov suffix. A separate output file is  also
72       created  for  each  file  included by the main source file, also with a
73       .gcov suffix added.
74
75       The basic content of an .gcov output file is a copy of the source  file
76       with  an  execution  count and line number prepended to every line. The
77       execution count is shown as - if a line does not contain any executable
78       code.  If  a  line  contains code but that code was never executed, the
79       count is displayed as #####.
80
81   OPTIONS
82       -a, --all-blocks
83              Display all basic blocks. If there are  multiple  blocks  for  a
84              single  line of source code, this option causes llvm-cov to show
85              the count for each block instead of just one count for  the  en‐
86              tire line.
87
88       -b, --branch-probabilities
89              Display conditional branch probabilities and a summary of branch
90              information.
91
92       -c, --branch-counts
93              Display branch counts instead of probabilities (requires -b).
94
95       -f, --function-summaries
96              Show a summary of coverage for each function instead of just one
97              summary for an entire source file.
98
99       --help Display available options (--help-hidden for more).
100
101       -l, --long-file-names
102              For coverage output of files included from the main source file,
103              add the main file name followed by ## as a prefix to the  output
104              file  names.  This can be combined with the --preserve-paths op‐
105              tion to use complete paths for both the main file  and  the  in‐
106              cluded file.
107
108       -n, --no-output
109              Do not output any .gcov files. Summary information is still dis‐
110              played.
111
112       -o=<DIR|FILE>, --object-directory=<DIR>, --object-file=<FILE>
113              Find objects in DIR or based on FILE's path. If  you  specify  a
114              particular  object file, the coverage data files are expected to
115              have the same base name with .gcno and .gcda extensions. If  you
116              specify  a  directory,  the files are expected in that directory
117              with the same base name as the source file.
118
119       -p, --preserve-paths
120              Preserve path components when naming the coverage output  files.
121              In  addition  to  the  source file name, include the directories
122              from the path to that file. The directories are  separate  by  #
123              characters,  with  .  directories removed and .. directories re‐
124              placed by ^ characters. When used with the --long-file-names op‐
125              tion,  this  applies to both the main file name and the included
126              file name.
127
128       -u, --unconditional-branches
129              Include  unconditional  branches   in   the   output   for   the
130              --branch-probabilities option.
131
132       -version
133              Display the version of llvm-cov.
134
135       -x, --hash-filenames
136              Use md5 hash of file name when naming the coverage output files.
137              The source file name will be suffixed by ## followed by MD5 hash
138              calculated for it.
139
140   EXIT STATUS
141       llvm-cov  gcov  returns 1 if it cannot read input files.  Otherwise, it
142       exits with zero.
143

SHOW COMMAND

145   SYNOPSIS
146       llvm-cov show [options] -instr-profile PROFILE  BIN  [-object  BIN,...]
147       [[-object BIN]] [SOURCES]
148
149   DESCRIPTION
150       The  llvm-cov  show command shows line by line coverage of the binaries
151       BIN,...  using the profile data PROFILE. It can optionally be  filtered
152       to only show the coverage for the files listed in SOURCES.
153
154       BIN  may  be  an  executable,  object file, dynamic library, or archive
155       (thin or otherwise).
156
157       To use llvm-cov show, you need a program that is compiled with  instru‐
158       mentation  to  emit  profile and coverage data. To build such a program
159       with clang  use  the  -fprofile-instr-generate  and  -fcoverage-mapping
160       flags.  If linking with the clang driver, pass -fprofile-instr-generate
161       to the link stage to make sure  the  necessary  runtime  libraries  are
162       linked in.
163
164       The  coverage  information is stored in the built executable or library
165       itself, and this is what you should pass to llvm-cov show as a BIN  ar‐
166       gument. The profile data is generated by running this instrumented pro‐
167       gram normally. When the program exits it will write out a  raw  profile
168       file,  typically  called  default.profraw,  which can be converted to a
169       format that is suitable for the PROFILE argument using  the  llvm-prof‐
170       data merge tool.
171
172   OPTIONS
173       -show-branches=<VIEW>
174              Show  coverage for branch conditions in terms of either count or
175              percentage.  The supported views are: "count", "percent".
176
177       -show-line-counts
178              Show the execution counts for each line. Defaults to  true,  un‐
179              less another -show option is used.
180
181       -show-expansions
182              Expand inclusions, such as preprocessor macros or textual inclu‐
183              sions, inline in the display of the  source  file.  Defaults  to
184              false.
185
186       -show-instantiations
187              For source regions that are instantiated multiple times, such as
188              templates in C++, show each instantiation separately as well  as
189              the combined summary.  Defaults to true.
190
191       -show-regions
192              Show  the execution counts for each region by displaying a caret
193              that points to the character where the region  starts.  Defaults
194              to false.
195
196       -show-line-counts-or-regions
197              Show the execution counts for each line if there is only one re‐
198              gion on the line, but show the individual regions if  there  are
199              multiple on the line.  Defaults to false.
200
201       -use-color
202              Enable or disable color output. By default this is autodetected.
203
204       -arch=[*NAMES*]
205              Specify  a  list of architectures such that the Nth entry in the
206              list corresponds to the Nth specified binary. If the covered ob‐
207              ject  is  a universal binary, this specifies the architecture to
208              use. It is an error to specify an architecture that is  not  in‐
209              cluded  in  the  universal binary or to use an architecture that
210              does not match a non-universal binary.
211
212       -name=<NAME>
213              Show code coverage only for functions with the given name.
214
215       -name-whitelist=<FILE>
216              Show code coverage only for functions listed in the given  file.
217              Each  line in the file should start with whitelist_fun:, immedi‐
218              ately followed by the name of the function to accept. This  name
219              can be a wildcard expression.
220
221       -name-regex=<PATTERN>
222              Show code coverage only for functions that match the given regu‐
223              lar expression.
224
225       -ignore-filename-regex=<PATTERN>
226              Skip source code files with file paths that match the given reg‐
227              ular expression.
228
229       -format=<FORMAT>
230              Use  the  specified  output  format.  The supported formats are:
231              "text", "html".
232
233       -tab-size=<TABSIZE>
234              Replace tabs with <TABSIZE> spaces when preparing reports.  Cur‐
235              rently, this is only supported for the html format.
236
237       -output-dir=PATH
238              Specify  a  directory to write coverage reports into. If the di‐
239              rectory does not exist, it is created.  When  used  in  function
240              view mode (i.e when -name or -name-regex are used to select spe‐
241              cific functions), the report is written to PATH/functions.EXTEN‐
242              SION.  When  used  in  file view mode, a report for each file is
243              written to PATH/REL_PATH_TO_FILE.EXTENSION.
244
245       -Xdemangler=<TOOL>|<TOOL-OPTION>
246              Specify a symbol demangler. This can be  used  to  make  reports
247              more human-readable. This option can be specified multiple times
248              to supply arguments to the demangler  (e.g  -Xdemangler  c++filt
249              -Xdemangler  -n  for  C++).  The demangler is expected to read a
250              newline-separated list of symbols from stdin and  write  a  new‐
251              line-separated list of the same length to stdout.
252
253       -num-threads=N, -j=N
254              Use  N threads to write file reports (only applicable when -out‐
255              put-dir is specified). When N=0, llvm-cov auto-detects an appro‐
256              priate number of threads to use. This is the default.
257
258       -line-coverage-gt=<N>
259              Show code coverage only for functions with line coverage greater
260              than the given threshold.
261
262       -line-coverage-lt=<N>
263              Show code coverage only for functions with  line  coverage  less
264              than the given threshold.
265
266       -region-coverage-gt=<N>
267              Show  code  coverage  only  for  functions  with region coverage
268              greater than the given threshold.
269
270       -region-coverage-lt=<N>
271              Show code coverage only for functions with region coverage  less
272              than the given threshold.
273
274       -path-equivalence=<from>,<to>
275              Map  the  paths in the coverage data to local source file paths.
276              This allows you to generate the coverage data  on  one  machine,
277              and  then use llvm-cov on a different machine where you have the
278              same files on a different path.
279

REPORT COMMAND

281   SYNOPSIS
282       llvm-cov report [options] -instr-profile PROFILE BIN [-object  BIN,...]
283       [[-object BIN]] [SOURCES]
284
285   DESCRIPTION
286       The  llvm-cov  report command displays a summary of the coverage of the
287       binaries BIN,... using the profile data PROFILE. It can  optionally  be
288       filtered to only show the coverage for the files listed in SOURCES.
289
290       BIN  may  be  an  executable,  object file, dynamic library, or archive
291       (thin or otherwise).
292
293       If no source files are provided, a summary line  is  printed  for  each
294       file  in the coverage data. If any files are provided, summaries can be
295       shown for each function in the listed files if the -show-functions  op‐
296       tion is enabled.
297
298       For  information on compiling programs for coverage and generating pro‐
299       file data, see SHOW COMMAND.
300
301   OPTIONS
302       -use-color[=VALUE]
303              Enable or disable color output. By default this is autodetected.
304
305       -arch=<name>
306              If the covered binary is a universal binary, select  the  archi‐
307              tecture  to use.  It is an error to specify an architecture that
308              is not included in the universal binary or to use  an  architec‐
309              ture that does not match a non-universal binary.
310
311       -show-branch-summary
312              Show statistics for all branch conditions. Defaults to true.
313
314       -show-functions
315              Show coverage summaries for each function. Defaults to false.
316
317       -show-instantiation-summary
318              Show  statistics  for  all  function instantiations. Defaults to
319              false.
320
321       -ignore-filename-regex=<PATTERN>
322              Skip source code files with file paths that match the given reg‐
323              ular expression.
324

EXPORT COMMAND

326   SYNOPSIS
327       llvm-cov  export [options] -instr-profile PROFILE BIN [-object BIN,...]
328       [[-object BIN]] [SOURCES]
329
330   DESCRIPTION
331       The llvm-cov export command  exports  coverage  data  of  the  binaries
332       BIN,...  using  the  profile  data PROFILE in either JSON or lcov trace
333       file format.
334
335       When exporting JSON, the regions, functions, branches, expansions,  and
336       summaries of the coverage data will be exported. When exporting an lcov
337       trace file, the line-based coverage,  branch  coverage,  and  summaries
338       will be exported.
339
340       The  exported data can optionally be filtered to only export the cover‐
341       age for the files listed in SOURCES.
342
343       For information on compiling programs for coverage and generating  pro‐
344       file data, see SHOW COMMAND.
345
346   OPTIONS
347       -arch=<name>
348              If  the  covered binary is a universal binary, select the archi‐
349              tecture to use.  It is an error to specify an architecture  that
350              is  not  included in the universal binary or to use an architec‐
351              ture that does not match a non-universal binary.
352
353       -format=<FORMAT>
354              Use the specified output  format.  The  supported  formats  are:
355              "text" (JSON), "lcov".
356
357       -summary-only
358              Export  only  summary  information for each file in the coverage
359              data. This mode will not export coverage information for smaller
360              units  such  as individual functions or regions. The result will
361              contain the same information as produced by the llvm-cov  report
362              command, but presented in JSON or lcov format rather than text.
363
364       -ignore-filename-regex=<PATTERN>
365              Skip source code files with file paths that match the given reg‐
366              ular expression.
367
368              -skip-expansions
369
370              Skip exporting macro expansion coverage data.
371
372              -skip-functions
373
374              Skip exporting per-function coverage data.
375
376              -num-threads=N, -j=N
377
378              Use N threads  to  export  coverage  data.  When  N=0,  llvm-cov
379              auto-detects  an  appropriate  number of threads to use. This is
380              the default.
381

AUTHOR

383       Maintained by the LLVM Team (https://llvm.org/).
384
386       2003-2023, LLVM Project
387
388
389
390
39112                                2023-07-20                       LLVM-COV(1)
Impressum