The committed patch fixes the issue only if an application redefines the
function names as described in README or call the prefixed functions directly.
But why making it more difficult, if the redefinition can be implemented in
perl2posix.h that must be included at any rate?
At first, I thought there is some benefit I cannot see and that dictates the
current status. But then I made some tests
<https://bugzilla.redhat.com/show_bug.cgi?id=1667614#c2> and came to a
conclusion that if an application includes pcre2posix.h, the only correct
operation mode is using PCRE's implementation.
Thus please consider this attached patch that simply by including
<pcre2posix.h> makes applications sure bad things won't happen.