Re: Row pattern recognition

Hi Henson, and hackers,

Henson has been mentoring me and suggested I try building and running
this patch, so this is a first-timer's build-and-use note, not a real
review.

I built cf/4460 (v49) on macOS 15 / Apple Silicon (arm64), meson with
cassert and debug. It built without errors. One small thing worth
flagging: it emits two -Wmacro-redefined warnings for INITIAL, in the
generated scanners (backend scan.c and ecpg pgc.c). The new INITIAL
keyword ends up as a token macro (#define INITIAL in gram.h) that
collides with flex's built-in INITIAL start condition (#define INITIAL 0
in the scanner). A plain master build here is warning-clean (no INITIAL
keyword there), so this looks specific to the patch. Seen on clang; not
sure if it's known or turns up on other compilers.

On tests, rpr, rpr_nfa, rpr_explain and rpr_integration pass. rpr_base
has only the RESPECT/IGNORE NULLS wording diff, which looks like the
pending rebase rather than anything real.

I also went beyond the basic example -- the two SKIP modes, quantifier
variants ({2,}, *, a reluctant quantifier, an alternation), and a few
odd inputs (tied values, a NULL mid-run, a single-row partition).
Everything behaved as I'd expect, nothing anomalous.

This was v49, since that's what cfbot has built. Happy to re-run on the
latest after a rebase, on macOS/arm64 if that platform is useful to have
covered.

Best regards,
Seongjun