Do you assemble slackware packages and install them or just do "make install"? Are git versions always stable? Are there compilation failures? Having a single script is of course more convenient than using a bunch of different slackbuilds from different (sometimes unknown) sources. If you compile everything yourself you don't need a binary distribution at all - you could use gentoo for example.

--enabled-shared isn't required technically. It should build with the static libs.

Yes, I do create slackware packages, I wrote Slackbuilds-like scripts for that. It's much better to make upgrades and potentially go back to a working packages if anything goes wrong.
But on the whole, I've been doing this for years and never had an unstable desktop yet. X11 and mesa packages are well developped, I almost never have compilation failure.
I actually wrote single Slackbuild for each package, and I ended writing a master script that starts the various Slackbuilds, with some error handling and logging. I guess I have some free time.
I just compile mesa drivers and some multimedia stuff, because they are fast changing, adding new features very often, but not releasing that often. For the rest of the OS, I'm quite happy with Slackware-current and alienBOB's kde and multilib packages. They're recent enough and compiling the whole thing would be tedious.
I don't think what I do is that much different to getting regularly new binary drivers. It's almost as easy since I've automated everything and not much longer.

True if you compile Mesa with "--disable-llvm-shared-libs". That's what I do. You get some warning during the compilation about that but I've never had any issue that could be tied to that. I don't do any gaming though, so I'm probably not pushing the OpenGL support to the limit.
It would be nice if Pat added shared libs to LLVM though.

Yes, it compiles, but acceleration(DRI) works after that(or falls back to sw mode)? I don't know for sure because I changed to shared libs before doing some other steps, but it seems to me - it should work, so step 3 from my post is not actually required if not upgrading llvm.

+1
And I'd also ask him to provide more recent mesa and other related stuff(glamor, etc). Why doesn't he do it even in current - maybe it's not considered so stable?

Im' running the GIT version of Mesa for a while now and I never experienced any instability. Games run even better with that. I don't think that's the problem, but maintaining a truckload of dependencies can give you a big headache.

Glamor will be included in xorg-server 1.16, which will be released in a week or so. Since mesa is working well - I'm going to see if I can screw it up and give the new xserver a whirl.

Good news! But when xorg 1.16 is integrated into slackware-current? For now we have only 1.14.x AFAIR.

When Pat thinks it is ready for inclusion. If you want it earlier you will have to build it yourself (the x11.Slackbuild makes that relatively easy).
For now it works to just use the external Glamor, works fine at least on my HD3200 and HD6870.

Yep. As far as I'm concerned, using an external library makes it easier to upgrade IMHO than upgrading the whole xorg-server, with a lot more potential side effects and dependencies.
But I guess it's easier for developpers to have it included in the server and maybe it'll increase adoption.