I really find it a poor show, Linus running his mouth like that, and all the others scurrying to reply deferentially and civilly to his adolescent rant (whether he is technically right or not, which doesn't even seem clear from the follow-up mails). IMHO, it's not the way to build an inclusive community, and it's not the way to encourage newcomers (who may be unfamiliar with your coding practices but could have useful expertise in their own right).<p>Really, i long for the "coolness" and "macho" around the Linux project to die. It's doing free software everywhere a disservice thanks to the project's high visibility. Maybe this makes me a wuss or whatever (i freely admit to being extremely non-confrontational), but it'd be nice if people of all cultures and communication styles would be welcomed as volunteers to great libre software projects such as these.<p>EDIT: I wonder if it would help if news sites would be more critical of such outbursts? This article seems to remain relatively neutral, but perhaps public condemnation of emotional outbursts would help towards taking the glamour out of it. I wonder if it'd also help stop "hero-worshipping" emulation of such behaviour.
Original ML mail: <a href="http://lkml.iu.edu/hypermail/linux/kernel/1510.3/02866.html" rel="nofollow">http://lkml.iu.edu/hypermail/linux/kernel/1510.3/02866.html</a><p>Someone rewrote the rant in polite English: <a href="http://catcode.com/comments/2015/cf20151101.html" rel="nofollow">http://catcode.com/comments/2015/cf20151101.html</a>
I urge all the people who haven't to actually read the mailing list post (link [0] from lobster_johnson), and note that (from the register article):<p>""" The rant is entirely impersonal: it rails against code, not people. Those who contributed the offending code will have no doubt of Torvalds' feelings towards it and the open nature of kernel development means it would not be hard to identify those responsible. Torvalds names no names, however. """<p>Linus is passionate about what he does, and as he has explained several times before, the reach and the medium make him easily misunderstood if he is all polite and politically correct, so he makes sure there is no way he is misunderstood.<p>Let me ask you this - after this issue, does anyone here think Linus will be willing to accept the usub()/uadd() calls into the kernel any time soon? Had Linus answered politely and politically correct, he would have had to do that about ten times as much, because (a) other people wouldn't notice or think that this response doesn't apply to their special snowflake code, and (b) those who know it applies to would feel that there is room for discussion.<p>Linus is herding cats without paying them, and has been doing this amazingly well for over 20 years now. Whether or not you subscribe to it, his management style works, produces amazing results. At the scale that linus manages, you (probably) have to be dictator, and (likely) cannot be a polite one.<p>[0] <a href="http://lkml.iu.edu/hypermail/linux/kernel/1510.3/02866.html" rel="nofollow">http://lkml.iu.edu/hypermail/linux/kernel/1510.3/02866.html</a>
Eventually Linux will find out that those new overflow functions produce better ASM code by checking the overflag FLAG. This is in HW, use it, and don't fallback to inline assembly.
jo +2 is much better than the previous code, even if the C function looks insane, yes. I agree with that. I also replaced all our crazy manual overflow checks with these new builtins, and with mult and signed add/sub it made it much faster and easier to read. The uadd/usub cases are indeed quirky, but at least consistent.<p>Summary: Linus is way off here.