Re: [lmi] PATCH: miscellaneous improvements to test_coding_rules

[Greg Chicares]

On 2021-06-17 21:06, Vadim Zeitlin wrote:
> On Wed, 16 Jun 2021 00:25:36 +0000 Greg Chicares <gchicares@sbcglobal.net> 
> wrote:
[...]
> GC> Okay, I'm ready for that whenever you like.

[...modulo twelve days...]

>  I've pushed this branch to misc-test_coding_rules in my repository,
> corresponding to https://github.com/let-me-illustrate/lmi/pull/187
> I won't write a long email about it because most commits should be
> self-explanatory

Yup.

I'll push this in a moment, with some changes of my own that ought
likewise to be self-explanatory.

I have some "BTW" comments about two commits, identified here by
their savannah SHA1s:

0f1bc0773 Split the check for canonical header guards in two parts

The old and new revisions behave identically AFAICS. Both test thus:

  // any lines at all
  opening include guard
  // any lines at all; next line must be the last:
  closing include guard

An enhancement would be to forbid anything interesting before the
opening include guard:

  // boilerplate C++ comments, blank lines: okay
  int xyzzy = 42; // NOT OKAY
  opening include guard [etc.]

in case you feel inspired to implement that (me, I'd rather return
to working on inverse quadratic interpolation).

30ba88acc Remove check for Latin-9 from coding rules test

I can now write
  6ϵ|ζ|
in comments, which is nicer than
  6.0 * std::numeric_limits<double>::epsilon() * std::fabs(zeta)
so this is a welcome change. I did worry about no longer filtering
out bizarre invisible characters that may occur in corporate
documents, but 'vim' makes them visible, e.g. as follows:
  Auf<200c>lage // zero-width non-joiner
  Lorem​<200b>Ipsum // zero-width space
  '<180e>᠎' // Mongolian vowel separator
so I guess there's nothing to worry about after all.

>, and while I do have some doubts about 0790ce22d (Move
> statistics display from test_coding_rules to makefile, 2021-06-02), I've
> already expressed them in the commit message itself.

I dropped that commit, because it changed the behavior:

original:
      692 source files
   197000 source lines
      276 marked defects

with that commit:
        1 source files
  6684988 lines
     1076 defects