You are not logged in.
Has someone considered some kind of automated build system for Arch?
Something that would work like this:
- It'd have every library and dependency possible installed.
- It'd intelligently read the makefiles produced to see what libraries they used and compare them against a lookup table to see what Arch dependencies they then required. Failing that, it could use ldd and a second lookup table that matched libraries to packages.
- It'd attempt to figure out the target binary to run (again, from the makefiles produced), and then run it. If it worked, it'd be marked as usable. If it didn't work, it'd be marked as needing fixing.
All of these points can fail, especially in the parsing of the makefiles; in each case, this would be noted by the system and user action could be taken.
In operation, it wouldn't take away from users managing their own packages. It'd just provide a secure environment to build packages in, and attempt to automate some of the process. In the best cases, the system would theoretically be capable enough to download a package's sourcecode, ./configure it, make it, make a package out of it, get the package verified as usable, then update the repo with it.
Note the verification step in the previous paragraph: I would never want this to be an automated system. Sure, it sound amazing on paper, and might even work for a little while, but sooner or later something would come crashing down and since repo management is quite a trust-based issue, everyone would freak out and they wouldn't want the build server anymore.
-dav7
Last edited by dav7 (2008-10-17 18:41:22)
Windows was made for looking at success from a distance through a wall of oversimplicity. Linux removes the wall, so you can just walk up to success and make it your own.
--
Reinventing the wheel is fun. You get to redefine pi.
Offline
Who would have access to upload to such a build server? If it's the general public, then this is a security nightmare, as well as a growth curve nightmare. The monetary investment for a project like this would need to come from somewhere.
And yeah, something like this has been considered, and a working proof-of-concept has been sitting around for years. http://projects.archlinux.org/?p=pacbuild.git;a=summary . What this kind of project really needs is someone with some distributed computing smarts and dedication (and time) to get it off the ground in a form that will survive past a proof-of-concept barebones implementation.
One of the largest design challenges would be dependency resolution for batch upgrades. For instance, let's say we update libfoobar, which is depended upon by foo, bar, baz, and batman, the system needs to know that libfoobar needs to be built and installed in order to compile the rest of them against it.
The suggestion box only accepts patches.
Offline