Am 23.04.2014 19:54, schrieb Jonathan Schleifer:
Am 23.04.2014 um 19:32 schrieb Ingo Weinhold <ingo_weinhold@xxxxxx>:We usually only vote when absolutely necessary, i.e. when a decision needs to be reached and a consensus cannot be found (respectively there isn't an option obviously favored by a majority). So, generally the process starts with an informal discussion. If a decision cannot be reached, a vote is suggested (the options may need to be discussed), and finally a formal vote is held (*). CU, Ingo (*) Years ago we almost formalized the process. There's mail from Jorge -- I'm too lazy to search it ATM -- which summarized the process and was generally agreed on. We should really put this stuff in the wiki or on the website.Ah, ok. I did not know that. So, we should first discuss about the pros and cons of Clang and then vote? I think it might sense to vote on it with an informal vote first, to see what the majority thinks. If e.g. everybody prefers GCC, that would make clear that it makes more sense to work on GCC 4.9.
A vote is still not the right tool. A vote is usually needed only when a discussion has shown that good arguments exists for multiple options, or that the pro and con arguments ballance for multiple options and a consensus can not be reached via discussion. A vote is like the last resort.
I think instead of gauging opinions via a "vote", which runs some risk to include a lot of uninformed or outdated opinion, you could lay out the pros of using clang.
For example, my little "knowledge" is that clang can be used as a service (like a library), so it integrates much better into an IDE. It also seems to provide much better errors and warnings, especially for template code, based on some examples I saw in a talk. And it appears to have a better overall architecture. I also heard that it can produce slower code than GCC. But this is all from a while ago and may, at least partially, be no longer true.
So while you seem to be pretty deep into the subject, you could give us your thoughts on clang versus GCC, and that may even be enough already to confirm a general plan to switch to clang once it's ready.
Best regards, -Stephan