Perhaps it’s made complex by searching for perfection when really you just need something that makes a high percentage of builds better.
You’d be crazy to create a final release of something without starting from a “clean” state, which means they DO NOT need a module implementation with 100% accuracy in all of C++’s asinine corner cases. I wouldn’t trust that accuracy even if they claimed it, I’d “clean” anyway.
This means that compilation speedup just has to ensure that MOST incremental rebuilds perform better without adding insane development overhead (such as having to respecify things).
If 60% of the time my incremental rebuilds are faster, I’d say that is more than enough to justify a C++1x release. They need to just move forward.
You’d be crazy to create a final release of something without starting from a “clean” state, which means they DO NOT need a module implementation with 100% accuracy in all of C++’s asinine corner cases. I wouldn’t trust that accuracy even if they claimed it, I’d “clean” anyway.
This means that compilation speedup just has to ensure that MOST incremental rebuilds perform better without adding insane development overhead (such as having to respecify things).
If 60% of the time my incremental rebuilds are faster, I’d say that is more than enough to justify a C++1x release. They need to just move forward.