Da
Show enters and exits. Hide enters and exits.
| 00:05:39 | boyscout | Fixed toplevel C ext tasks to always delegate to the ext Rakefile. - 60f7d84 - Brian Ford |
| 00:08:31 | boyscout | CI: 60f7d84 success. 3002 files, 11431 examples, 35502 expectations, 0 failures, 0 errors |
| 01:07:29 | rue | brixen: Ouch 60f7d84. What is the current thinking for the new build process? |
| 01:07:45 | brixen | ouch? |
| 01:10:02 | rue | Sympathetic "ouch" ;) A change is clearly needed |
| 01:10:55 | brixen | yeah, definitely |
| 01:11:04 | brixen | so, there are a couple fundamental problems with rake |
| 01:12:04 | brixen | there is a single global dependency tree is the biggest |
| 01:12:31 | brixen | anyway, the goal is to create a dependency lib |
| 01:13:01 | brixen | and create instances af an ExtensionBuilder that has its own dependency graph |
| 01:13:15 | Zoxc | and don't forget Windows compatibility |
| 01:13:17 | Zoxc | ;D |
| 01:13:25 | brixen | and participates in the build process by responding to messages like :configure, :build, :clean |
| 01:13:30 | brixen | Zoxc: exactly the point |
| 01:14:11 | brixen | the build process would essentially be an aggregate of these instances for each thing |
| 01:14:29 | brixen | building a C ext, building an external library, building the VM, building the kernel, etc |
| 01:14:34 | rue | Lib to be driven from Rake, or an independent beast |
| 01:14:40 | rue | +? |
| 01:14:43 | brixen | probably independent |
| 01:14:54 | brixen | rake is horrid for cmd line args |
| 01:15:01 | brixen | so I don't see a reason to use it |
| 01:31:20 | rue | So main division: 1. creating rbx and 2. adding to rbx; creation subdivided in prerequisites, vm, kernel and then, essentially, file management |
| 01:32:59 | rue | Which further breaks into dependency management, various native build methods, build option handling corollary to which is platform detection |
| 01:41:00 | brixen | jcrosby_: http://groups.google.com/group/fa.caml/browse_thread/thread/3b1bb6f2da3ef270# |
| 01:41:22 | brixen | rue: basically, yeah |
| 01:53:31 | rue | Probably not worth it to go highly modular or generic |
| 01:55:32 | brixen | my model is essentially to use distinct classes for the various tasks |
| 01:55:49 | brixen | similar to how DependencyGrapher and InstructionParser are done |
| 01:56:05 | brixen | ExtensionBuilder, LibraryBuilder, etc |
| 01:56:31 | brixen | these would simply understand a protocal for stringing them together to complete full or partial builds |
| 01:56:43 | brixen | the actual work is well encapsulated |
| 01:56:59 | brixen | and each can use dependency graphs internally to define file targets |
| 06:26:20 | rue | Hm |
| 08:21:35 | tarcieri | yawn |
| 08:54:07 | rue | Indeed |