https://bugs.exim.org/show_bug.cgi?id=2769
--- Comment #11 from Carlo Marcelo Arenas Belón <carenas@???> ---
(In reply to Thomas Tempelmann from comment #10)
> The runtime check is very fast, and it'll only be done when I compile a
> regex for JIT. Besides, It's understood that compiling a regex takes time,
> whereas then performing the search through the JIT compiled code is fast,
> and presumably doesn't even run into this check then, right? So, I see
> absolutely no problem here.
you will need to do at least 2 calls to this function per compilation, and
might need to call it an even number of times at processing time as well if
your JIT is somehow self modifying.
> There is no way around this runtime check IF one wants to support older
> macOS versions.
could you elaborate on why you can't target 11 for Apple Silicon and whichever
older version you would like to support in Intel?
the test I reported in comment #4 did that and used `lipo` to generate a
(AFAIK) working version of a universal binary.
> The only thing you could do is to remove the runtime check IF the code is
> built for a minimum of macOS 11. But since this runtime check does not cost
> any measurable performance, this is overkill.
would this patch on top allow you to complete your build, which versions are
you planning to support and test it on?
https://github.com/carenas/sljit/commit/accdee8ed0331f525d18186fffc4c69c46374fe8
--
You are receiving this mail because:
You are on the CC list for the bug.
