Change all post-increments to pre-increments, where it doesn't affect the result #1135

Postincrement (var++) requires making a temporary copy of var in C++.

Wow many code changes for a little win, but I'm fine with them.

Binary changes to: addrman, bitcoinrpc, db, guiutil, keystore, main, net, script, wallet, walletdb

http://luke.dashjr.org/tmp/code/preinc.bin.diff.bz2

I consider pre-increment/post-increment more of a code style preference - no need to change it if it doesn't gain you anything.

I consider unnecessary post-increments to be akin to unused variables.

@sipa +100

The number of changes could be brought down if you'd only do this for non-primitive types (ie, iterators).

But even then, this is a kind of a micro-optimization that makes sense for tight inner loops. I'm not sure it's worth changing everywhere.

FWIW, KDE fixed this in 2007.

I just compiled a simple for-loop twice (i++ and ++i) and compared the disassembled code. There were no differences even when using -O0.

This isn't the only cleanup for 0.6.1 that has almost no effect on binary output. I suspect the actual binary differences are all in the iterators.

I've just talk to someone who knows a thing about C++ compilers. He said iterators from the std library can be inlined and thus it++ vs ++it won't make a difference there either.

In case we overload ++ in Bitcoin this might in fact make a difference but the better way to fix it would be by making the overloaded method inline-able.

That's assuming glibc and GCC. Best to do things the "right way" so we don't depend on implementation-specific optimizations.

Too much code change for zero gain.