=== maclin1 is now known as maclin | ||
=== maclin1 is now known as maclin | ||
=== kickinz1\afk is now known as kickinz1 | ||
=== maclin1 is now known as maclin | ||
=== bdmurray_ is now known as bdmurray | ||
cyphermox | infinity: hello sir. | 18:17 |
---|---|---|
infinity | cyphermox: Guten Tag. | 18:18 |
cyphermox | infinity: there's skiboot in trusty queue, if you have time for reviewing things, please :) | 18:18 |
cyphermox | and ^^ ibmpmlinux for partner. | 18:18 |
infinity | cyphermox: I can haz skiboot for vivid too? I assume these are straight backports from wily? | 18:20 |
cyphermox | yes, we should haz skiboot on vivid too... | 18:20 |
=== gsilva is now known as gsilvapt | ||
infinity | cyphermox: ^-- Both accepted. | 18:36 |
infinity | That sure built fast... | 18:39 |
infinity | cyphermox: Hrm. Looking at the build log, it seems this package intends/wants to build big-endian, but since you didn't build-dep on a cross-compiler, it's building LE. Was that intentional, or an oops? | 18:43 |
cyphermox | I think that's intentional, and it seemed to work from my testing on two different machines... but given that I have very little knowledge on how precisely it's supposed to work, it's hard to say if I missed something | 18:45 |
infinity | cyphermox: Well, it'll limit the usefulness a bit, but that depends on IBM's usecases. Most (all?) PPC firmware is big-endian intentionally, because that works on all hardware and emulated environments, while LE is a crap-shoot of "it might work, sometimes". | 18:46 |
cyphermox | fwiw, I didn't make the initial packaging, I expect there was a conscious decision to build little-endian too | 18:46 |
cyphermox | wait | 18:46 |
cyphermox | this doesn't build the firmwares though | 18:46 |
cyphermox | just opal-prd. | 18:46 |
infinity | cyphermox: I'm not sure if it was conscious effort, or an accident. You have to stare at the logs to realise it's *trying* to cross, then falling back to native when it can't. | 18:46 |
infinity | cyphermox: Oh, this isn't the skiboot firmware bits, just a userspace tool? | 18:46 |
cyphermox | it's all from the same large source, but for now it just builds the opal-prd userspace tool | 18:47 |
infinity | cyphermox: Okay. So, the userspace tool should absolutely be native-endian, while the firmware shouldn't be. Their build system seems to conflate the two. | 18:47 |
cyphermox | hmm | 18:47 |
infinity | cyphermox: If and when we get around to building both from the same source, we should decouple those. | 18:47 |
infinity | cyphermox: So that firmware can be BE and userspace is native-endian of current userspace (and respect cross, if crossing). | 18:48 |
infinity | cyphermox: But not an issue for today. | 18:48 |
cyphermox | OK | 18:48 |
infinity | Firmware should also be freestanding, which can be BE even without a cross toolchain. | 18:48 |
infinity | Because gcc is smart that way. | 18:48 |
infinity | (See grub2) | 18:48 |
infinity | So, if you're spending real time on this one with upstream, suggest they fix that up, but this is fine for now. | 18:49 |
cyphermox | alright | 18:49 |
* infinity accepts the binaries. | 18:49 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!