| stekern | die patch queue, die! ;) | 06:41 |
|---|---|---|
| stekern | would have been fairly annoying to apply the patches to both orpsocv2 and or1200 "joined", so hope no-one minds that I did them as seperate commits | 06:44 |
| stekern | (I have orpsocv2 and or1200 as seperate git-svn checkouts) | 06:44 |
| olofk | stekern: Referring to the PIC patches? | 08:58 |
| stekern | no, the or1200 patches | 08:59 |
| olofk_x2000 | stekern: Good job with digging up the old patches | 11:36 |
| stekern | is that an olofk on a train or an olofk cloned 2000 times? | 12:46 |
| * juliusb hopes it's olof cloned 2000 times | 12:52 | |
| stekern | orpsocv3 will be done in no time! ;) | 12:54 |
| -!- olofk_x2000 is now known as olofk_x1999 | 12:56 | |
| olofk_x1999 | Damn it. The clones aren't stable | 12:56 |
| stekern | haha | 13:11 |
| juliusb | stekern: where were you going to commit your dynamic linking changes? to the github or1k-gcc/src repo right? | 13:13 |
| juliusb | I'm updating the opencores GNU tool chain wiki page with links and information about the development version | 13:14 |
| stekern | yes, to the openrisc/or1k-src repo | 13:16 |
| stekern | gcc wasn't in the patches I posted, it will come in time | 13:18 |
| juliusb | ah right | 13:18 |
| juliusb | nps | 13:18 |
| juliusb | I'm adding you and pgavin to the tool chain development team | 13:18 |
| juliusb | you've both contributed in epic porportions | 13:19 |
| jeremybennett | juliusb: stekern: Do you know where the or1k-src repo takes its upstream from? | 13:27 |
| juliusb | (actually, by adding you to the "team" I meant "list of the team") | 13:31 |
| stekern | jeremybennett: I believe it's the sourceware cvs, but I'm mightly confused by that tree, so I'm very thankful to Peter who put together the or1k-src tree ;) | 13:37 |
| stekern | actually, it's not as confusing as I remember, looks like it's 1:1 copy of it | 13:39 |
| jeremybennett | stekern: Thanks | 13:49 |
| juliusb | stekern: I've put on the wiki some basic instructions on building the development version fo the tool chain, based on the one you pastebinned me a while back | 13:50 |
| juliusb | I found myself wanting to do it again | 13:51 |
| juliusb | http://opencores.org/or1k/OpenRISC_GNU_tool_chain#Installation_of_development_versions | 13:51 |
| stekern | nice, surely a better place to keep them than on pastebin ;) | 13:58 |
| juliusb | agreed :) | 13:59 |
| stekern | I have similiar notes for the linux toolchain, I've used /opt/or1k as install path for them (in contrast to /opt/or1k-toolchain), which sounds better? | 14:00 |
| stekern | I need to figure out what's broken with the -muclibc -mnewlib flags in gcc | 14:01 |
| juliusb | I have just done it in /opt/or1k for my work now | 14:11 |
| stekern | ok, should we change the wiki instructions for or1k-elf to /opt/or1k then? | 14:21 |
| stekern | I've changed it now | 14:26 |
| stekern | bikeshed issue, but we seemed to agree on the color ;) | 14:27 |
| jeremybennett | stekern: Thanks for doing this. You would be better building with a unified source tree (i.e. both or1k-src and or1k-gcc together). Then you'll only need one GCC build pass. | 14:41 |
| jeremybennett | As an example look at https://github.com/foss-for-synopsys-dwc-arc-processors/toolchain | 14:41 |
| jeremybennett | It's not quite the same - we don't have a live upstream sourceware mirror, so we use the upstream mirrors of the individual components, but it may have some useful ideas for you. | 14:42 |
| jeremybennett | It also builds a uClibc tool chain, based on Mike Frysinger's instructions, which may be useful for you. | 14:43 |
| stekern | but... how will I keep my basement warm, if I don't have to let the cpu spend some cycles on building the second stage gcc? ;) | 14:45 |
| stekern | more seriously; personally, I like building them seperately, it made me understand better the seperation between gcc and the sourceware tree | 14:49 |
| stekern | besides, someone might not be interested in building gcc at all | 14:50 |
| stekern | and is only interested in installing binutils | 14:51 |
| -!- olof_ is now known as olofk_x1 | 15:25 | |
| stekern | so... is that the original olof, or is it a last remaining clone? | 15:34 |
| jeremybennett | stekern: I understand about separate building. However if you don't wish to, GCC's build system understands about newlib and can build in one pass. | 15:48 |
| stekern | jeremybennett: yes, I'm not denying it's useful information, I appreciate it | 15:49 |
| stekern | is that the case for uClibc too? or is that always necessary to build in 2 steps? | 15:49 |
| jeremybennett | No - newlib is a special case. uClibc needs multiple passes. Mike Frysinger came up with a standard way to do this (I think as part of the Gentoo project), which is what we use in the ARC example. It is roughly the same approach used in bld-all.sh | 15:50 |
| jeremybennett | potentially you need more than two steps for complete bootstrap, but we can usually get away with two. | 15:51 |
| olofk_x1 | stekern: I haven't found out yet if I am myself, or if I am trapped in a virtual machine | 16:23 |
| olof_ | Off to Sue Ellen! | 20:01 |
| juliusb | olofk: Sue Ellen, I'm jealous :) That was a good place | 20:35 |
| juliusb | olofk: sorry, I didn't follow the FSCONS stuff too closely, did you email them a proposal? | 20:35 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!