On 2023-01-12 21:54, Carlo Arenas wrote:
> On Thu, Jan 12, 2023 at 7:38 PM Paul Eggert <eggert@cs.ucla.edu> wrote:

> Fair enough, this will likely need a new test though, and of course
> changes to the current ones as well.
> 
> Right now they will report that grep doesn't have '-P' support instead
> of reporting that unicode is missing, and take into consideration
> those tests that set multibyte locale were successful after my change,
> so they will also need changes as they would misbehave silently
> otherwise.

OK, I installed it with the attached additional patch, which attempts to 
address these issues. The new test is done by init.cfg each time you run 
require_pcre_ in a UTF-8 locale. The current tests are changed by 
altering require_pcre_ to do the right thing in the en_US.UTF-8 locale 
(I also tightened up the other locales' testing slightly). The new 
require_pcre_ uses a different diagnostic when checking en_US.UTF-8 
locale. Since all the PCRE tests are done in either the C or the 
en_US.UTF-8 locale, this should be good enough.