Clang: Defending C++ from Murphy's Million Monkeys | Hacker News Reader

Clang: Defending C++ from Murphy's Million Monkeys

channel9.msdn.com

95 pointschmike16 comments

As a reference, this video was posted 3 weeks ago with another title in http://news.ycombinator.com/item?id=3565400. But almost receive no upvote.

Yeah, you need some luck to make it to the front page. If I recall correctly, one vote can be enough, but it has to be done soon after submission time.

Once you get past that threshold, karma skyrockets.

The analogy to start-up success is startling.

Taken from that perspective -- thank god for re-submissions.

It gets worse because not many "interesting" people dedicate some time to /newest. I used to go there once every day. The main problem is HN doesn't discriminate by topic combined with lack of downvote. The noise to signal ratio is too high. There are way too many spam, blogspam and metoo articles against a few actual interesting submissions. And most times a blogspam article gets to the frontpage instead of the original source. It's a shame.

In reddit the people doing the honorable job of upvoting new stories are called "Knights of New". This effect is notable on good communities like /r/programming. Trash submissions quickly get downvoted.

I have been using clang for just over a year now and I have found that without it I start going insane. It is by far my favourite compiler. The error messages it spits out are sane (compared to those of GCC) and help me track down the problem faster and with less effort. It definitely has made me a very happy man!

Does anyone have a clue as to why GCC's messages have remained so bad for so long?

gcc's error messages, especially for C++ templates, got much better in 4.5. They don't have pretty colors, but I didn't find a substantial benefit from clang 2.9's vs gcc 4.5's for C++ development when using both for a week.

Note that if you're on a Mac, your system's gcc is likely several years out of date.

Even with 4.5 I haven't found the error messages to be nearly as helpful. Also with clang you can easily turn on a way to look for undefined behavior and turn those into warnings/errors which is fantastic and helped me track down a bug BEFORE it caused me huge grief.

> Note that if you're on a Mac, your system's gcc is likely several years out of date.

To be precise: Apple stopped maintaining its GCC branch after the switch to GPL3. The last version available with Apple patches is 4.2.1. There will be no further updates to this branch.

Check out [Colorgcc](http://schlueters.de/colorgcc.html). It helps more than you'd think.

Because it's hard to add it to the mature infrastructure that was not designed to support it from the beginning. Clang was built from ground up to support good diagnostics.

I use Clang everyday and the error messages, warning messages, and static analysis are really great to work with.

Good talk minus the bashing of GCC and Stallman. It's a free product you don't need to use it and you're not entitled to anything as the license clearly states. The speaker pretends to be against open source politics but doesn't state his bias since both Apple and Google have corporate interests to undermine GPL compiler tools.

On the other hand, I really like the competition. It's very healthy to have options.

If clang takes off, GCC will take a lot more bashing, simply because meaningless flame-wars are the nastiest.

If there's a clear winner, you run with it. If the tools are really different, (i.e. Python vs. C), you pick the right tool for the job. If the tools do essentially the same thing in slightly different ways, with slight advantages and disadvantages in some areas (i.e. KDE / Gnome; vi / emacs; Linux / BSD) then you resort to ad hominem attacks.

So much wisdom in so few words, this was one of the times where just an upvote wasnt enough.

I don't believe it was GCC bashing but rather exposing an unintended consequence of a political decision that led to clang. Sad but true.

Reply on news.ycombinator.com