A question if I may, leading on from
http://bugs.freepascal.org/view.php?id=25440

- inc(x)/dec(x) with range checking enabled are "optimized" with taking address of x.

Is that why Inc() and Dec() can't be applied to properties, or are there other issues?

They were using untyped address, causing loss of alignment and generating large and slow code with unaligned access. This is fixed in r26256 and improves all alignment-sensitive targets.

Some platforms, for example ARM/Linux and Solaris/SPARC, fix alignment traps in the kernel so programmers might be unaware of any problems. Unfortunately, they might also be unaware that this introduces a substantial performance hit.

Incidentally, I notice that somebody in Russia is still developing SPARC systems http://lists.debian.org/debian-sparc/2013/12/msg00006.html http://en.wikipedia.org/wiki/MCST-R1000 I wonder whether they're aware that the architecture is well-supported by FPC and Lazarus?

http://wiki.freepascal.org/User_Changes_2.4.0#Treating_direct-mapped_properties_as_regular_fields

Regards,
Sven

Noted, but I think my real question is whether the compiler could recognise that the optimisation is inappropriate and instead generate +=n (with range checking).

Frankly, I'm fairly happy with the += idiom but I appreciate that there are some people who object that "it's not Pascal". I think that an alternative to having to repeat the full name of a deep property is desirable: it's obviously possible to use with but some people don't like that either :-)

Yes, the compiler was recognizing case with enabled range checking and generating "x:=x+n". However if "x" involves a function call (e.g. inc(somearray[foo()+2]); ) it is still necessary to avoid calling it twice, hence the optimization. I used quotes because the end result was opposite to the expected one.

Regards,
Sergei

While this behavior is absolutely logical, as a normal procedure could not use a property in that way.

OTOH just as a courtsey, the _builtin_ procedures inc() and dec() could be implemented in a way that allows for doing it.

-Michael