> it is more convenient that the whole testing doesn't stop by an
> error of a single test file.

> If we don't have to keep the original behavior of "make check",
> let's modify the rule of ".el.elc:".

So the test code byte compilation rule would not stop Make, but would
be captured in the .log file and reported at the end summary report?
That is fine to me.

We should have ert-summary-report kill-emacs with positive error code
if there were such errors. That way the overall 'make check' will
still fail if there were build errors.