kcov loses accumulated data when ${1:?} used

[Mark-E-Hamilton]

Hi, again,
I've got another simple test case.

#!/bin/bash
#echo before
fparam=${1:?}
echo $fparam

When I run it with a parameter I get 100% coverage

% rm -rf kcov_out
% kcov ./kcov_out ./test7.sh something
something

When I run it with no parameter I get 0% coverage.

% rm -rf kcov_out
% kcov ./kcov_out ./test7.sh
./test7.sh: line 3: 1: parameter null or not set

Howwever, when I run them both in sequence I get 100% coverage after the first kcov, and 0% after the second.

% rm -rf kcov_out
% kcov ./kcov_out ./test7.sh
case 1
% kcov ./kcov_out ./test7.sh
./test7.sh: line 3: 1: parameter null or not set

FInally, if I uncomment the 'echo before' statement I get 100% coverage in both cases when I run them in the order above. If I run them in the other order (the failure case first) I get 33% coverage after the failure case, and then 100% coverage after the success case (which seems correct.)

[SimonKagstrom]

OK, this is sort of an interesting case. Apparently bash will parse this as a syntax error and return an error code:

ska@localhost:kcov$ bash -x tests/bash/dollar-one-question.sh
tests/bash/dollar-one-question.sh: line 3: 1: parameter null or not set
ska@localhost:kcov$ echo $?
1

adding a parameter sets that right again:

ska@localhost:kcov$ bash -x tests/bash/dollar-one-question.sh 4
+ fparam=4
+ echo 4
4
ska@localhost:kcov$ echo $?
0

and in this case, kcov works as expected. However, as you noted, the real kcov issue here is that accumulation doesn't work as expected. I have to investigate, but my guess is that the reason is that kcov visits files in bash "lazily" the first time they are referenced through PS4. So in the error-case, kcov won't see any files at all. When you added the echo, the file will show up and the accumulation can be done.

[Mark-E-Hamilton]

Thanks.

I suppose then that I'll have to visit my larger test case. I strip out
code from large scripts to get the smallest possible piece that still
exhibits the problems I see. Occasionally I add statements that might
not have been in the original, so my test cases can be too simple. In
the larger case I'm seeing a similar problem, but your description makes
me think it might actually be something else. I'll have to investigate
it some more.

---

Mark E. Hamilton
Engineering Sciences Center
Senior Member of Technical Staff
Sandia National Laboratories
505-844-7666

[SimonKagstrom]

The smaller tests are quite good to have, so thanks for doing that!

Is your original problem an accumulation problem?

[Mark-E-Hamilton]

I think so. When I run a single test case the shell lines in question
are covered. When I run all the test cases the previously covered lines
turn red again. I haven't been able to isolate it yet, however.

---

Mark E. Hamilton
Engineering Sciences Center
Senior Member of Technical Staff
Sandia National Laboratories
505-844-7666