IRC logs for #openrisc Monday, 2015-01-12

--- Log opened Mon Jan 12 00:00:52 2015
-!- FreezingAlt is now known as FreezingCold04:24
olofkSaw someone on the opencores forum who complained of a missing ___umodsi3. That was fixed quite recently, wasn't it?08:58
stekernolofk: no, I don't think there ever been a problem with it10:40
stekernok, so there has been bugs fixed that relates to umodsi3, but it was never missing10:56
stekernI answered the forum guy10:56
olofkcool. I didn't really understand the issue on IRC. Just remembered something about umodsi10:56
stekernit's good that you do, it could have been related10:58
olofk_franck__: Happy birthday!11:33
olofkstekern: You don't happen to have an avalon 32 to 64 bit converter?11:34
stekernno, but doesn't altera have a bunch you can use?11:39
olofkah.. right11:39
stekernI think if you are using qsys and connect a 64-bit bus to a 32-bit bus, it'll insert converters automatically11:39
olofkThat could be an option. Feels like my setup is complicated enough already11:41
stekernI bet you can poke them out manually in some mega-something though11:43
olofkIf only my DDR2 controller worked reliably :/11:48
olofkHow do I start qsys?11:48
olofkstekern: There's some qsys support in fusesoc, right?12:16
_franck__olofk: thanks. Who betrayed me ? :)12:27
olofk_franck__: LinkedIn :)12:30
stekernolofk: yup, sure is12:44
olofkstekern: Ok, so I've created a qsys thingie now. What next?12:58
olofkShould I generate verilog files somehow?12:58
stekernyou add the qsys file to qsys_files under [quartus]13:07
stekernthe when you run fusesoc build, it will generate the verilog files for you in the build dir13:08
olofkIIRC you had some manual step as well. Was that because of some special setup in your case?13:10
stekernno, that was before the qsys support was added to fusesoc13:11
olofkah ok13:11
olofkIt always takes something like 10 seconds for emacs to start up. There must be something weird set up. Any tips for debugging?13:23
amylaarblueCmd: We at Embecosm wanted to have a look at the regression test failures of the OpenRISC port.  So first we need to reproduce this.  Is the gcc in branch gcc-trunk?14:57
stekernamylaar: it's
blueCmdthe other repo is the staged repo for mainline inclusion. its most probably out of date16:44
aburgessstekern, blueCmd in the or1k-gcc repo, is it the or1k branch that is the interesting one?16:52
stekernaburgess: yes, that should be the most recent16:54
aburgessstekern, thank you16:54
stekernlatest sync was against the 4.9.1 release though, so it's slightly dated in terms of following mainline16:54
olofkHas anyone tried to build newlib from upstream?17:33
olofkHaven't seen wallento around lately17:33
olofkHmm. I must be doing something wrong when I install the newlib toolchain19:04
olofkI'm getting a lot of undefined references19:04
olofkand multiple definitions19:04
olofk/opt/or1k/lib/gcc/or1k-elf/4.9.1/../../../../or1k-elf/lib/libor1k.a(syscalls.o): In function `_write_r':19:04
olofk/home/olof/code/bld/newlib/or1k-elf/libgloss/or1k/../../../../../newlib-2.2.0/libgloss/or1k/syscalls.c:32: multiple definition of `_write_r'19:05
olofkThat wasn't a good paste19:05
olofkThis is when I compile a hello world with or1k-elf-gcc19:06
olofkOr is the upstream newlib broken?19:20
olofkBased on previous experience, my first guess is that the fault is on my side19:21
ysangkokpoke53282: do you know if gdb would work in or1ksim?19:28
poke53282ysangkok: Yes, I know the answer. ;)20:32
poke53282The answer is the same for jor1k.20:33
ysangkokah :( ok20:33
poke53282we need better ptrace capabilities of the kernel and further changes in gdb for the  openrisc platform.20:35
poke53282ptrace works already. But only after one weekend of fiddling with the kernel and ptrace.20:36
poke53282strace works too.20:37
poke53282sorry, I mean ltrace20:38
poke53282ltrace works already. But only after one weekend of fiddling with ptrace in the kernel and ltrace.20:39
poke53282I know it is confusing.20:39
poke53282strace, ltrace, ptrace.20:39
ysangkokyeah :P20:46
-!- Netsplit *.net <-> *.split quits: LoneTech21:14
ysangkokpoke53282: so, i wanted to add the ability to specify fs metadata in the query string, to enable loading thirdparty CORS content (i believe the "src" field is already an url?). but to make it even easier to inject random data, it would be useful if there were a was to specify file content in the filesystem json. this would make it easier to test different changes without changes the urls which are point21:47
ysangkoked to.21:47
ysangkokpoke53282: i have two solutions that i think are both ok: make the "src" field a URI instead, which means one could use Data URI's with optional base64 encoding21:48
ysangkokpoke53282: the other solution is to add a new field and mandate the base64 encoding. this frees us from implementing a custom data URI parser (i believe this is also not available to javascript)21:49
ysangkokpoke53282: do you have any thoughts on this matter?21:49
poke53282the src field is a link to a file, if the normal way to determine the link fails.22:45
poke53282like files which begin with . or _.22:45
poke53282Of course, this field can be extended for uri.22:46
poke53282but only as an option.22:46
poke53282both should be possible.22:47
poke53282But I don't understand the base64 encoding. Why?22:47
ysangkokpoke53282: because it's how people usually put binary data in json?23:19
--- Log closed Tue Jan 13 00:00:53 2015

Generated by 2.15.2 by Marius Gedminas - find it at!