1) Get this repo file: http://repos.fedorapeople.org/repos/spot/chromium-stable/fedora-chromium-stable.repo and put it in /etc/yum.repos.d/
2) EDIT - I've signed the packages with my personal GPG key, upon request. This means you also need to download my public key. You can either get it from:
or by running:
gpg --recv-key 93054260 ; gpg --export --armor 93054260 > spot.gpg
Then, copy it (as root) to /etc/pki/rpm-gpg/
cp -a spot.gpg /etc/pki/rpm-gpg/
3) Run yum install chromium chromium-v8
Why has it been so long?
A) I really believe in building things from source code. At the very least, I should be able to take the source code and know that I can follow instructions (whether in an elaborate README or in src.rpm) and get a predictable binary that matches it. This is one of the core reasons that I got involved with FOSS oh so many years ago.
B) I'm a bit of a perfectionist. I wanted any Chromium builds to be as fully functional as possible, and not just put something out there with disabled features.
C) Chromium's not my day job. I don't get paid to do it by Red Hat (they don't mind me doing it as long as I also do the work they _do_ pay me for).
D) I also build Chromium using as many system libraries as possible, because I really get annoyed at Google's policy of bundling half the planet. I also pull out and package independently some components that are reasonably independent (webrtc, ffmpegsumo).
If my schedule is reasonably clear, it takes me about 7-10 _days_ to get a Chromium build up and going, with all of its components. Why did it take a year this time around? Here's the specifics:
AA) I changed jobs in November, and that didn't leave me very much time to do anything else. It also didn't leave me very motivated to drill down into Chromium for quite some time. I've gotten my feet back underneath me though, so I made some time to revisit the problem....
BB) ... and the core problem was the toolchains that Chromium requires. Chromium relies upon two independent (sortof) toolchains to build its Native Client (NaCl) and Portable Native Client (PNaCl) support. NaCl is binutils/gcc/newlib (it's also glibc and other things, but Chromium doesn't need those so I don't build them), and PNaCl is binutils/llvm/clang/a whole host of other libs. NaCl was reasonably easy to figure out how to package, even if you do have to do a bootstrap pass through gcc to do it, but for a very long time, I had no success getting PNaCl to build whatsoever. I tried teasing it apart into its components, but while it depends on the NaCl toolchain to be built, it also builds and uses incompatible versions of libraries that conflict with that toolchain. Eventually, I tried just building it from the giant "here is all the PNaCl source in one git checkout" that Google loosely documents, but it never worked (and it kept trying to download pre-built NaCl binaries to build itself which I didn't want to use).
** deep breath **
After a few months not looking at Chromium or NaCl or PNaCl, I revisited it with fresh eyes and brain. Roland McGrath was very helpful in giving me advice and feedback as to where I was going wrong with my efforts, and I finally managed a building PNaCl package. It's not done the way I'd want it to be (it uses the giant git checkout of all PNaCl sources instead of breaking it out into components), but it is built entirely from source and it uses my NaCl RPMs. The next hurdle was the build system inside Chromium. The last time I'd done a build, I used gyp to generate Makefiles, because, for all of make's eccentricities, it is the devil we understand. I bet you can guess the next part... Makefile generation no longer works. Someone reported a bug on it, and Google's response is paraphrased as "people use that? we should disable it." They've moved the build tool over to something called "ninja", which was written by Google for Chromium. It's not the worst tool ever, but it's new, and learning new things takes time. Make packages, test packages, build, repeat. Namespace off the v8 that chromium needs into a chromium-v8 package that doesn't conflict with the v8 in Fedora proper that node.js uses. Discover that Google has made changes to namespace zlib (to be fair, its the same hack Firefox uses), so we have to use their bundled copy. Discover that Google has added code assuming that icu is bundled and no longer works with the system copy. Discover that Google's fork of libprotobuf is not compatible with the system copy, but the API looks identical, so it builds against the system copy but does not work properly (and coredumps when you try to setup sync). Add all the missing files that need to go into the package (there is no "make install" equivalent).
Then, we test. Discover that NaCl/PNaCl sortof works, but nothing graphical does. Figure out that we need to enable the "Override software rendering list" in chrome://flags because intel graphics are blacklisted on Linux (it works fine once I do that, at least on my Thinkpad T440s, your mileage may vary). Test WebRTC (seems to work). Push packages and hope for the best. Wait for the inevitable bugs to roll in.
I didn't do an i686 build (but some of the libraries that are likely to be multilib on an x86_64 system are present in i686 builds as well), because I'm reasonably sure there are not very many chromium users for that arch. If I'm wrong, let me know. I also haven't built for any older targets.