Re: traveling for a few days

[Branden Robinson]

Hi folks,

[I'm still email-compromised]

On Sun, Apr 24, 2022, 08:39 Bjarni Ingi Gislason <bjarniig@rhi.hi.is> wrote:

> On Sun, Apr 24, 2022 at 04:39:57AM +1000, John Gardner wrote:
> > On macOS, Groff builds and installs perfectly from a fresh checkout
> > (currently using f55d8f41).
> >
> > However, running `make check` produced 6 failing tests:
> >
> > FAIL: src/devices/grotty/tests/basic_latin_glyphs_map_correctly.sh
> [...]
>
>   The bug was reported in ticket #61891 (closed), but was not made
> robust (invariant about the charmap containing an utf-8 encoding)
>

I did _attempt_ to make it thus; see
https://git.savannah.gnu.org/cgit/groff.git/commit/?id=4115620b4d10d0d2b1d45eb9ceca2825e742579d

for the code changes that attempt to skip the check in question. I'm
curious to know how John's environment is interacting with the code added
there.

  The test fails if the user's charmap is not as prescribed by the
> test.
>

The test should be _skipped_ if the locale's charmap is not UTF-8.

  It suffices to have installed any locale with an UTF-8 charmap and
> using that, for example
>
> environ="LC_CTYPE=$(locale -a | grep -i -e 'utf.*8' | head -n 1)"
>
>   and using "env $environ" in front of
>
> "$grotty" ...
>
> and
>
> od -t c
>

In my opinion, it complicates the test too much to grep through the system
for some--any--UTF-8-using locale and then using it.  Such a process is too
non-deterministic for automated testing--some systems will lack such a
locale altogether.

Run the test suite itself using a UTF-8 locale, e.g., with
"LC_ALL=en_US.UTF-8 make check" if maximal test coverage is desired. If
it's not, the test should be SKIPped, but should not FAIL.

I think the objective for now should be to discern why John's environment
is not triggering the skip logic shown at the URL above.

Regards,
Branden