IRC logs for #openrisc Wednesday, 2015-10-14

--- Log opened Wed Oct 14 00:00:25 2015
stekernI agree, generators are great for some tasks. But, they must be transparent.04:50
andrzejr_franck_, barebox works but I had to disable caches (otherwise it gets stuck in cache_init) and ethernet (no ethernet in hw yet)07:16
andrzejrwill use that testcase to debug the cache more thoroughly07:17
andrzejralso, there was a compilation error: arch/openrisc/cpu/cpu.c:44:58: error: ‘RESET_SCOPE_CPU’ undeclared (first use in this function)07:18
olofkI like code generators as well for some tasks, but as others have said, you need to understand what they do, and they should create understandable code.08:30
mafmblueCmd_: assuming that the last contributor of GCC decides to assign the copyright to the FSF, what's next?10:31
mafmis the port of GCC for o1rk up to date?10:31
olofkandrzejr: I should probably prepare a blog post for my ideas of FuseSoC IP-Xact integration. Not sure when I'll find time to do that though10:34
stekernmafm: blueCmd_ merged the gcc 5.2.0 release into our port, so we're at least that much up to date10:42
mafmstekern: uh, that's good :)  I was fearing that it needed more porting, since gcc-5 looked like a release with huge changes overall (but perhaps in the front-ends, not back-ends)10:45
stekernhe mentioned that the merge was a pain ;)10:46
mafmno kidding10:53
olofkAnyone handy with busybox?10:53
mafmso I suppose that if the copyright assignement is sorted out, the Debian port can be continue right where it was left10:54
olofkRight now I just want to understand where the defconfig is locaed10:54
olofkAlright. Managed to compile busybox-1.24.0 with musl by using the defconfig and disabling wtmp, utmp and rpc manually10:59
olofkIs there any easy way to test my newly compiled busybox?11:01
olofkI guess I could just replace the one in my initramfs and give it a go11:02
poke53281olofk: clone jor1k, exchange busybox in sys/or1k/fs/bin/11:03
olofkpoke53281: That's more like it. How do I run it? Do I need a local webserver?11:03
poke53281unfortunately yes. Or you have to disable the security of the web browsers.11:04
poke53281I think in python a web server is a one liner or so.11:04
poke53281the sys directory is a submodule. So not just clone but also "git submodule init" I guess.11:05
olofkI'm trying with or1ksim to begin with. Got that installed already11:05
olofkBut I should definitely learn how to fire up a jor1k instance. That could come in handy11:05
olofkwhoops. Crashed hard11:06
poke53281I think you have to type three lines in bash and then open a web browser.11:06
poke53281git clone ....11:06
poke53281cd ...11:07
poke53281git submodule ....11:07
poke53281python -m SimpleHTTPServer11:07
poke53281Ok, these are four lines11:07
poke53281I definitely like the filesystem_as_hardware_device thingy. It makes things much much easier.11:09
olofkI'll take a look at that sometime11:09
olofkRight now my ordinary kernel crashes as well, so there is something strange here11:09
olofkoh.. wrong device tree11:11
olofkAlright! Looks like my busybox works11:12
olofkdalias: FYI it looks like busybox has fixed the broken kernel includes in
olofkI'm using upstream linux headers 3.15 instead of those from sabotage linux now11:15
olofk...I think :)11:15
olofkAny way to check that?11:15
olofkDo we have a busybox defconfig for openrisc btw?11:20
poke53281Do you use musl?11:20
olofkyes. My end goal is to rebuild the initramfs in our kernel tree, so that we can use an upstream kernel and build an initramfs ourselves11:21
olofkWith musl11:21
poke53281scripts/progs.make gives you the build instructions11:22
olofkcool. I'll diff your config with the default one11:23
olofkI see that UTMP and WTMP is set in our config11:26
olofkAnd you're using sysroot when you compile11:26
poke53281Hmm, don't really understand.11:28
poke53281Yes, I have a sysroot directory.11:28
poke53281Building a distribution is still a horrible mess. Especially now, that I have to mix glibc, musl, openrisc and risc-v.11:32
olofkaha... looks like I need to set either PATH_WTMP or WTMP_FILE as well. Ehhm.. what should I set that to?11:33
poke53281don't know11:33
olofkIt's not set in your config. Maybe it's a new thing11:33
poke53281might be11:34
olofkHmm.. it's not set anywhere in the default config for 1.24.0, so I can't really see how this can work at all11:35
olofkMaybe the C library is supposed to set that11:36
olofkoh well. I'll leave that for now11:38
olofkIs there any automated way to create all the busybox symlinks and root fs layout?11:38
stekernolofk: make install11:41
olofkWhere does it install to?11:46
olofkI probably shouldn't try it as root :)11:46
stekernto the sysroot11:46
olofkoh, so that's what sysroot is good for11:46
olofkSo should I point sysroot to the arch/openrisc/support/initramfs before building busybox?11:47
stekernor to your sysroot11:48
stekernand then just copy everything in there11:48
olofkSo my sysroot can be an empty dir?11:48
stekernwell, the one I'm using is at: /opt/cross/or1k-linux-musl/or1k-linux-musl11:53
olofkBut I don't want my busybox to be installed there, right?11:53
stekernyeah, maybe not11:56
stekernto be honest, I can't remember what I did last when I compiled busybox with musl11:56
olofkaha. There's a menuconfig option for this :) (./_install) BusyBox installation prefix11:57
poke53281olofk: arghh, sorry olof. The file I send you is outdated. I forgot all the musl patches11:58
poke53281One second11:58
poke53281I changed the busybox scripts11:59
olofkThis link looks a bit like the one I just downloaded :)12:00
poke53281Yes, but the file content changed12:00
olofkInternet is so sneaky12:00
olofkSo where should I look?12:00
poke53281you have a new patches/busybox folder12:00
olofkJesus. That's a lot of patches12:00
poke53281look in  script/progs.make and search for busybox12:01
poke53281Yes, musl and busybox doesn't fit very well together.12:01
poke53281dalias will tell you that.12:01
olofkAre these compile-time or runtime fixes? Because with latest busybox, it compiles and starts up fine by just disabling the three options I mentioned earlier12:02
poke53281good question. I guess both.12:02
olofkAnd make CROSS_COMPILE=... install works fine12:02
poke53281If it runs don't worry. I had problems with an older busybox. But I can't remember which problems.12:03
olofkWould be interesting to get some more testing on this. Maybe most of those patches aren't needed anymore12:03
olofkYeah. I had a lot of problems compiling 1.23.212:03
poke53281ask dalias and the sabotage linux guy12:03
olofk#musl on freenode?12:04
olofkHaven't there been several problems recently that was solved by turning of the caches? Do we have a potential problem here?13:13
poke53281which cache? I have never used or implemented a cache.13:25
olofkpoke53281: You might not know this, but there even exists implementations of or1k that can run on an FPGA ;)13:29
poke53281Yeah, I have a rumor in that direction one or two times.13:32
poke53281... have heard ...13:35
-!- Netsplit *.net <-> *.split quits: rhythmx, mboehnert, yang, GeneralStupid, simoncook16:50
-!- mboehnert1 is now known as mboehnert16:50
-!- Netsplit over, joins: simoncook17:07
jeffesquivelsHi guys, any idea if the talks at Orconf were recorded? (I realize if they were they probably won't be readily uploaded right now, but just wondering if they are going to be available in the future)18:11
jn__i'm interested, too18:11
olofkjeffesquivels, jn__ : Actually, I think they are quite close to being available. Apparently they are edited and just waiting to be uploaded20:23
jeffesquivelsolofk: great to hear that! I'll be on the lookout then, thanks! :)20:30
--- Log closed Thu Oct 15 00:00:27 2015

Generated by 2.15.2 by Marius Gedminas - find it at!