Measuring executed shell scripts coverage under shunit

[vlajos]

Hi,

Referencing this: http://stackoverflow.com/questions/27555639

Recipe to reproduce:
git clone git://github.com/vlajos/misspell_fixer.git
cd misspell_fixer
git checkout -qf 2d16e17c7901bd06e24d67f315e782f148d45d0b
curl -L "https://shunit2.googlecode.com/files/shunit2-2.1.6.tgz" | tar zx
rm -rf coverage;/usr/local/bin/kcov coverage/ test/tests.sh

And there isn't anything in the output regarding the misspell_fixer.sh

Finally I rewrote most of my script this way:
vlajos/misspell-fixer@b5ec1f7

Thank you.

Best regards,
Lajos

[SimonKagstrom]

I tried running it with the lastest kcov commit, and I get coverage output at least. The above commit sees to it that the misspell_fixer.sh output to stderr is shown.

I saw you mentioned that set -x breaks coverage collection. This is true, since kcov actually uses the regular bash trace functionality to collect coverage data. I'm not sure if there's some workaround to fix that particular issue. There is a BASH_XTRACEFD environment variable, which perhaps can be used in some way, but I'm not sure it really bites.

[vlajos]

I just compiled the latest version and used my above script, but I see only coverage report about tests.sh and shunit2 and nothing about the "main" misspell_fixer.sh.
Do I miss something?

[SimonKagstrom]

I tested the example you gave above (with the git commit referenced), and I get misspell_fixer.sh coverage. I ran it like

kcov /tmp/kcov ./misspell_fixer.sh -v ~/projects/kcov

and all coverage I see is from misspell_fixer.sh. How do you execute your test?

[vlajos]

This way:
/usr/local/bin/kcov coverage/ test/tests.sh

[SimonKagstrom]

OK, good. I can reproduce it with that.

[SimonKagstrom]

Fixed with 57f6959.

The problem with the current solution was that kcov starts bash with -x <script>, which doesn't propagate into other programs executed. The fix above instead adds a file to be read in via BASH_ENV, so then it should be set for all scripts.

[vlajos]

Thank you!

[vlajos]

I don't know if it is related or not, but the loop_core function isn't covered . (The same recipe above.)

It is called this way: (after L200)
find "$@" -type f $cmd_part_ignore -and ( $opt_name_filter ) -exec bash -c 'loop_core "$0"' {} ;

The testReal0NoBackup test should call it.
The warning "Starting script" and warning "Done." before/after the find are covered.

[SimonKagstrom]

OK, I saw that too. I guess it might be a similar issue.

I can reproduce it with your script.