Hi again,

"I test the code on different hosts, since it's depends on the environment."

That's one thing, and should be done for sure. But generating a test coverage report depends on the codebase, not the environment, unless you're using it in some way I've not encountered.

But generating a test coverage report depends on the codebase, not the environment, unless you're using it in some way I've not encountered.

I hadn't chimed in because I've never tried merging the coverage report. But I can think of lots of situations where the environment will influence which code can be covered in a given environment, and thus will influence the test coverage report.

Here are three examples:

# the coverage report on MSWin32 will be different on Win10 from Ubunt
+u Linux.  
# I assume the OP would like to merge the two reports to make sure tha
+t all the 
# functions are covered at least once between the two (or more) differ
+ent 
# environments
if($^O eq 'MSWin32') {
  run_this_function();
} else {
  run_a_different_function();
}

# or

# the coverage report is different in the environment where 
# MY_ENV_VAR is set true vs where it's false or undefined,
# and the OP would like to merge the two reports to make sure that all
+ the 
# functions are covered at least once between the two (or more) differ
+ent 
# environments 
if( $ENV{MY_ENV_VAR} ) {
  run_special_case();
}

# different parts of the code are covered depending on whether some ot
+her 
# executable is found, so that it can be run
if( -x '/x/y/bin/my_executable' ) {
  my $ret = qx(/x/y/bin/my_executable); # cannot run on system X, wher
+e repo is in /a/b/repo/run.pl, and /x/y/ hierarchy might not even exi
+st!
  ... error processing ...
  $some_outer_variable += $ret;
}
[download]

In all of these situations, on system X the coverage will show certain functions or blocks of code run, but others not; and system Y may have a different group of functions or blocks that are run. So for example, maybe each system can run the same 80% base coverage, but system X covers 5% of code blocks that cannot run on Y and system Y has 15% of code blocks that cannot run on X; the individual reports say 85% coverage on X and 95% coverage on Y, when in reality, together they make 100% coverage, but neither report shows that. So the OP is asking how to merge those so that they can tell whether or not all lines of code are covered in at least one of the test systems.