Output/UI should not rely on colors

[cyrille-artho]

Right now, when running benchexec, the output of each verdict ("true" or "false") is shown in green or red in order to indicate whether this is a true or false verdict.
The output should not only rely on color additionally use a different way of marking the correctness of the verdict, e.g., "!" vs. "?", or bold (if the terminal supports it).
This will accommodate red/green color blindness and also avoid confusion in cultures where the meaning of red/green is swapped by default (e.g., in Japan).

[PhilippWendler]

Thanks for bringing this up.

First, I have to say that the text output of BenchExec is primarily intended as showing progress to the user, and potentially also to let them see if things are going completely wrong (like everything immediately producing an error) and they might want to cancel the whole run in order to not waste resources. I hope users are not under the impression that they have to rely on this (quite basic) output in any way for actually looking at the results.
Additionally, even in the text output of BenchExec we provide a summary of correct/wrong results, so even users for whom the color does not work will be able to see whether everything was correct or how many errors there were. And for finding out which tasks produced certain results we have the tables.
So I see the colors as a nice-to-have feature for when it works, but nothing (should) rely on it and BenchExec is usable without them (you can actually turn them off with the environment variable NO_COLOR=1 even).

But of course, that doesn't prevent us from polishing the colors if there is something that can be done.

Regarding bold: Actually all colors (red, green, orange, magenta) are already bold in the output, only results with category missing (no property or expected verdict available) are not bold and have no color set. I guess we could also "unbold" the green output. Then errors, wrong, and unknown results will be bold as a way to highlight "this might need your attention" to the user.

I would prefer to not add markers like ? or !. From user questions, I have the impression that the category is not always immediately understandable already, and using textual markers only in one specific place but nowhere else would further complicate that.

What do you think?