Branch coverage is reported wrongly

[theold190]

In some cases you might have 0 branches in all files. In that case branch coverage is reported differently in a HTML report:

• per file it is reported as 100%, 0/0
• per module it is reported as exec:0, total:0, coverage: 0%

It looks like 0% value is selected to avoid the zero division. An extra check could be added to show a proper statistics information.

Tested with official gcovr 3.2 release.

[davis-tim]

I can confirm the different behaviour.

The calculation for the branch coverage for the coverage report is:

    branches_covered = 100.0 if class_branches == 0 else 100.0*class_branch_hits/class_branches

But for the module the coverage is calculated as follows:

    coverage = 0.0 if branchTotal == 0 else round(100.0*branchCovered / branchTotal,1)

[mayeut]

Also related, #149

A unique behavior shall be chosen for this case. I'd like to see something other than choosing between 0 % and 100 % if possible (PR ready if you're interested):
A light gray background and - % string

e.g.

[latk]

@mayeut That looks like a reasonable approach to visualize this case. If you already have a PR ready do submit it please. I would have otherwise deferred this issue until after some improvements from the dev branch have been backported (refactoring into multiple files, and template-based HTML reports).