On Sun, Aug 23, 2026 at 1:22 PM Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx> wrote:
>
> You're making an incorrect assumption about evaluation order.
To add a little more color to this: With HIST_SUBST_PATTERN the
evaluation order changes, "s/b/$rep" does become "s/b/_". This is
handled at the level of the actual replacements, not at the point
where $str:s/b/$rep is initially parsed.