[17:12:33] <Krinkle>	 OK, I'll see what I can do. I'll work on a patch first to confirm the findings and make sure I can reliably reproduce it on a latest master and self-compiled.
[17:13:12] <Krinkle>	 I'm already starting to see it isn't as straight forward. When compiling it relatively plainly out of the box on latest 8.5.0-dev, the issue isn't there on macOS.
[17:13:19] <Krinkle>	 `./configure --enable-debug --enable-intl --enable-mbregex --enable-mbstring --with-external-pcre --with-iconv=/opt/homebrew/opt/libiconv --with-libedit && make -j8`
[17:13:30] <Krinkle>	 That's what I'm using so far.
[17:13:43] <Krinkle>	 Compared to https://github.com/Homebrew/homebrew-core/blob/7a7679884bcce97cc378e733417631bcb02e2678/Formula/p/php.rb which is the 8.4 package that does reproduce the issue.
[17:14:16] <Krinkle>	 I'll try checking out 8.4 to confirm it happens there and if not, narrow down more of the flags passed there to see what combo is required to trigger it.
[17:30:38] <Krinkle>	 Actually I'm not sure I will. Besides these three extensions I can't think of why any others would influence this. I'll probably dig into the pcre code paths and figure out what it's doing. It would have to other not be making tables for some reason, or making them correctly meaning isspace varies behaviour for reasons other than setlocale
[17:32:19] <Krinkle>	 I did look at the Apple OSD repos for when libc/isspace changed behaviour but couldn't splot anything obvious. There is a major update that pulls in a newer version of FreeBSD/libc and I've looked in their repo for any intentional changes there, but haven't found it there either.