Comment # 7 on bug 1183960 from Cristian Rodr���guez

(In reply to Michael Matz from comment #6)
> An even more aggressive solution would be to stop having the pcre2-posix
> library at all.  Sources that consciously want the pcre2 variant of the
> POSIX regex facility can include pcre2posix.h without other changes in their
> sources and link
> against libpcre2 (without -posix).  And sources who do not consciously want
> pcre2
> under the POSIX names wouldn't link against libpcre2-posix anyway.  So
> providing
> libpcre2-posix at all seems to be a mere rope waiting to be hanged by.
> 
> But if the above is deemed too aggressive, then yeah, at least moving it to
> a separate package, not pulled in by mere chance, and making sure that no
> libraries
> we provide link against it, would be a solution as well.  So, thanks for
> doing
> that :-)  (I came from the gdb side via ncurses to look at this, and my
> initial reaction was: ncurses linking against pcre2-posix?  ugh, disaster in
> waiting :) )

I did take the semi-axe solution, indeed having *any shared library* linked
against libpcre posix is a bad idea and something that needs to be wholly
banned and allowed by whitelist special case only as it will infect everything
up the chain.