Hi Przemek,
On Fri, 30 May 2008, Szakáts Viktor wrote:
It's not a joke but very serious argument. I even said
the same few times on this forum.
Sometimes it's not worth to make some things very complicated
to safe few CPU cycles because other person (or even the
same after few months/years) may have to update the code and
lost time for analyzing the exceptions and looking for possible
bugs.
For me, it caused about 5-10% of performance loss when compiling
the full BCC build using non-GNU make files. This 10% percent
is _5 seconds_ (+-2 because I used a very non-scientific method
to measure). And this is with a 5 years old PC (P4 2.6/800 HT 3GB RAM).
So, my vote is to dump WIN32_LEAN_AND_MEAN, before the release.
BTW in HVM we have code covered by HB_MACRO_STATEMENTS which does
not work for very long time and probably should be removed too or
at least updated to work with current code.
For HB_MACRO_STATEMENTS my vote is to remove it ASAP, even
before the release, if possible.
I'm all for removing all minor (major) variations and build
time options. All of them just add up to the complexity and
exponentially raise testing / support effort.
Brgds,
Viktor
_______________________________________________
Harbour mailing list
[email protected]
http://lists.harbour-project.org/mailman/listinfo/harbour
