| --- Log opened Sat Apr 02 00:00:02 2016 | ||
| mithro | I can't seem to get or1k gcc to compile - following the instructions at http://openrisc.io/newlib/building.html | 00:52 |
|---|---|---|
| mithro | It fails on newlib with "compiler can't create executables" | 00:53 |
| mithro | Oh, I see - I cloned the newlib directory into the gcc directory | 01:10 |
| wallento | systemgotyou: do you mean OpenRISC-based SoC? | 02:46 |
| olofk | wallento: re git checkout. Ah, so before every time you copy stuff to the build directory, you checkout the revision you want. That's pretty clever | 08:44 |
| olofk | We would need to apply patches on the fly then, but that's doable | 08:46 |
| olofk | Could probably make it backwards-compatible by just adding some options to the git/github providers | 08:47 |
| olofk | Note however that the github provider isn't using git at all. It's more a special case of the url provider | 08:48 |
| olofk | But all this is more of an optimization. It's still possible to just download the whole repo for every core, so I think other things should be prioritized for now | 08:51 |
| olofk | wallento, andrzejr: I'm starting to run out of reasons to not use providers as generators, so I'm getting ready to pull in andrzejr's patches to begin with | 09:39 |
| olofk | Some things need to be addressed though, but that could be done later | 09:40 |
| olofk | 1. Need to let cores register themselves as providers, so we don't have to put everything into FuseSoC | 09:40 |
| olofk | 2. Need to invalidate the cache, when the input files changes | 09:41 |
| olofk | The second item needs to be done for the cache anyway, so it's not strictly tied to providers | 09:41 |
| olofk | I've been thinking before about just checksumming or checking timestamps of the input files and store that somewhere | 09:42 |
| olofk | That is also something I want to do with the build tree, so we don't have to clean it out every time | 09:42 |
| olofk | But it needs to be done in a smart way, so we don't just end up reimplementing make :) | 09:43 |
| wallento | olofk, you actually don't need to check out the entire git to one revision | 09:54 |
| wallento | git checkout <revision> -- <files> checks out only the provided files from revision | 09:55 |
| wallento | as we have a given fileset that should be okay | 09:55 |
| wallento | we can then apply patches in tree and mv | 09:55 |
| wallento | moving ensures we can reuse the same repo for other cores if necessary | 09:56 |
| wallento | otherwise conflicts can happen | 09:56 |
| andrzejr | olofk, thank you. Generator providers felt like a hack to me initially but they work really well. IMHO they are no different from built-in support for simulators or synthesizers. | 13:19 |
| andrzejr | having user-defined plugins for fusesoc could be handy indeed. Not just for providers. | 13:20 |
| olofk | wallento: I like the idea. We should at least plan that for the future | 14:34 |
| olofk | andrzejr: Yeah, I feel the same. It's great that you came up with the idea and has tested it for a while too | 14:35 |
| olofk | And plugins is good, I agree. Just don't feel like defining API for it all right now. This needs to be pretty stable | 14:36 |
| olofk | andrzejr: Could you rebase your changes against FuseSoC master? I've picked some of your patches, but not all | 14:49 |
| olofk | ahh... my work switched vpn solution to use l2tp over ipsec or something like that. What a fucking mess | 18:21 |
| olofk | Anyone handy with that? I gave up now | 18:22 |
| --- Log closed Sun Apr 03 00:00:04 2016 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!