OpenMandriva Is Finding Great Success In Their Switch To Using LLVM's Clang Compiler

[aht0]

I'm not sure why you people seem so invested in programs or having unquestionable control over them, but for people in the business world who just want to make a tool, use a tool and get stuff done, it's not like fucking someone's wife, it's a business decision. I don't have any problem with publishing something under BSD/MIT and neither do major corporations like IBM and Microsoft. Go ahead and use my code, contribute if you want, or don't, I don't care, I'm not running a GNU/Cult.

Personally I'm just eager for a Linux without anything GNU so people will stop harassing strangers on how they should refer to it. Was never one for politics, that's why I avoid twitter as well. I don't like being harassed for simply what I'm thinking.

It's a question of reverse-morality. The very same GNU die-hards would have no compunctions on "stealing" (as they are putting it) the non-GPL code, so they naturally assume any other person would to the same. It does not matter that devs licensing the software chose permissive license knowingly. They just cannot really comprehend anyone having different set of base values than them. Mommy teached them"stealing is baaad", so in order to get rid of morally "questionable" things, they hate them with a burning passion.

[berolinux]

If compiling a single file with two compilers the elapsed time is max(gcc, clang), not gcc+clang. This is because the two can run in parallel on a multi-core CPU.

Unless of course all cores are already at the limit because stuff is built with "make -j$(getconf _NPROCESSORS_ONLN)" or the likes -- which happens to be done in pretty much all of our packages. On top of that, our build machines usually build several packages at the same time already.

The box I'm on right now is an 8-core (16 threads) Ryzen 2700x. All cores are at 100% load, Load average 46.64, 37.33, 22.47. Compiling things twice would certainly slow things down (and ccache wouldn't help a lot unless there was a giant disk for caching stuff -- it's not the same package being built all the time).

[skeevy420]

I agree. The cache size should be at least several gigabytes. I am using several caches of size 2-4 GiB.

The cache hit ratio for Gentoo package builds is currently at 31% on my machine. The hit ratio for $HOME/.ccache is 90%. The hit ratio depends on which build system the developer is using.

A long term hit ratio of X% means that Load average on your machine could drop by up to X%.

The age of some files in the cache can be several months.

Enabling ccache unconditionally for all builds has a small overhead in the worst case scenario.

There is CCACHE_COMPRESS env variable which enables compression. There is also CCACHE_COMPRESSLEVEL. Compression can be enabled in ccache.conf as well.

Thanks. I learned something of value today.

[mastermind]

Read this free book to put the GPL vs permissive debate to rest https://mises.org/library/against-in...ual-property-0

[cybertraveler]

Read this free book to put the GPL vs permissive debate to rest https://mises.org/library/against-in...ual-property-0

I used to be a huge fan of the GPL. It was reading things like the above that changed my mind about it. The principles outlined in that book and others like it are sound. Copyright and Patents are anti-property. Their enforcement involves actual property rights violations.

It's somewhat ironic that Stallman is pro "Free Software" and not fond of "Open Source" on the basis that Free Software is supposedly about freedom and the principles whereas Open Source is more about pragmatism. I later discovered it's actually the other way around.

However, I don't see the GPL as the work of the devil. I think Stallman & Co have done great things & with good intentions. His work has lifted us up. I think we just need to keep climbing higher.