Finish PlainStatusBackend

[ralismark]

#487 implements most of this but appears abandoned. This fixes some of the comments left on those changes.

[codecov]

Codecov Report

Merging #636 into master will increase coverage by 0.31%.
The diff coverage is 28.73%.

@@            Coverage Diff             @@
##           master     #636      +/-   ##
==========================================
+ Coverage   45.94%   46.25%   +0.31%     
==========================================
  Files         138      139       +1     
  Lines       60587    60448     -139     
==========================================
+ Hits        27834    27958     +124     
+ Misses      32753    32490     -263     

Impacted Files	Coverage Δ
src/bin/tectonic.rs	0.00% <0.00%> (ø)
src/status/mod.rs	59.25% <0.00%> (+0.63%)	⬆️
src/status/plain.rs	0.00% <0.00%> (ø)
src/status/termcolor.rs	37.37% <0.00%> (-1.58%)	⬇️
src/driver.rs	60.50% <70.00%> (-1.70%)	⬇️
tests/executable.rs	89.79% <90.00%> (+66.91%)	⬆️
src/errors.rs	6.38% <0.00%> (+0.02%)	⬆️
... and 13 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1a200dd...0dbaf2a. Read the comment docs.

[pkgw]

Thanks again for your contribution! I especially appreciate your working from @Mrmaxmeier's branch to as to preserve their commits in the history.

With your update my only comment is about the test coverage. It would be nice to add something there to keep the coverage moving in the right direction. I think if you add a case or two to test/executable.rs that use the new color=no option, that should do the trick?

[pkgw]

Ahhh ... I was confused by why you new tests didn't seem to have any coverage, but here it is disabled in the coverage run!

Looking in the history, it is not clear to me why these tests get disabled during the code-coverage tests. @Mrmaxmeier I think you added these tests originally ... do you remember why?

[Mrmaxmeier]

I vaguely remember something about the coverage tracking not working for end-to-end tests that spawn a new tectonic process. Not sure if that's still an issue though.

[pkgw]

I've pushed some commits that will try to re-enabled the executable tests in the coverage run. I'm not sure if they will correctly add to the coverage stats (as per @Mrmaxmeier's comment), but hopefully they'll at least run.

[pkgw]

OK I think it's the case that the executable tests are running, but the coverage induced by executing the tectonic program is not being recorded. I'll do a bit of research to if there's anything that we can do about that ... hopefully someone has tackled this problem before.

If that doesn't work out we can certainly go ahead and merge this, but I think it's worth some effort trying to solve this.

[pkgw]

xref: assert-rs/assert_cmd#9

[pkgw]

OK, I did a little experiment. If I hack up the executable test harness to manually launch its tectonic subprocesses inside kcov, I can make it so that we do gather coverage information from those invocations. Doing so gains us about 1.3% in code coverage (~1000 lines) compared to the current baseline. I think I could wire this up in the CI/CD framework, but it would be a bit hacky and ugly.

Alternatively, we could work on revising the executable-type tests to use Tectonic as a library, with an absolutely minimal amount of code in src/main.rs (or its moral equivalent).

Or, we could keep those tests as-is and just keep in mind that they won't contribute to the code coverage stats.

I am not sure which of these options I prefer. Any opinions?

[pkgw]

OK, well, I am still interested in any thoughts about this broader question, but work along those lines is best saved for a separate PR. So, let's merge this one!