IRC logs for #openrisc Saturday, 2013-01-19

stekernblueCmd: the regression tests went by with no new regressions, so I've pulled in your patches now09:27
blueCmdstekern: cool, did it fix anything? :P10:15
stekerngcc and g++ tests where exactly the same10:59
stekernlibstdc++ tests have always failed to even start, they seem to be working now10:59
stekern....but some test ran out of mem so it didn't run to the end11:01
stekernI'm going to increase the memory on it and rerun11:01
stekernblueCmd: here's the results if you want to take a look:
blueCmdstekern: cool11:05
stekerngcc/gcc.log gcc/gcc.sum and g++/g++.log g++/g++.sum are the files that are of most interest11:05
blueCmdhah, omg the number of tests11:06
stekernyeah, 47/78324 isn't that bad after all ;)11:07
blueCmdhehe, i guess.11:07
stekerntarget is of course 011:10
blueCmdstekern: you don't happen to know if there has been any discussion about using/not using foo@X syntax for GAS?11:15
blueCmdI'm working on TLS for openrisc, and I need to introduce a couple of relocs, and right not i'm thinking about using ".long foo@tlsgd"11:16
stekernI used the plt(foo), got(foo) just because that was easiest to implement11:33
stekernin contrary to foo@plt11:35
blueCmdah, I see. I will try that11:36
blueCmdbecause this @ is quickly turining to be quite intrusive11:36
blueCmdstekern: I'm trying to regenerate the cgen opcodes. I've read README-OR1K but I got stuck at this part:
stekernblueCmd: oh, we ought to remove that README-OR1K... it's outdated...13:37
stekernjust configure or1k-src with --enable-cgen-maint13:37
stekernand the cgen stuff will be regenerated when you make13:37
stekernsorry bout the confusion :(13:38
blueCmdnp :)13:38
blueCmdglobal offset table thread pointer offset high - gottpoffhi13:39
blueCmdnameing things is hard :P13:39
* Xark links to a handy utility -> :-)13:45
blueCmdhaha, Ghaskelang is a much better name - I agree13:45
XarkHey, worked for Tolkien. :)13:46
blueCmdl.lwz     r3,gottpofflo(foo)(r10)13:49
blueCmdthat doesn't look neat13:50
XarkIt is going to be tough to come up with an intuitive name for that... make sure to add a comment on its definition. :)13:58
XarkMaybe an underscore or two would help (if it fits with the style)...13:59
blueCmdheh, well it's called gottpoff in almost every other arch so I will stick to that13:59
XarkExcellent reasoning. :)13:59
blueCmdI wanted to do hi(gottpoff(foo)) first, but that wouldn't go with the other (gotoffhi / gotofflo) already existing ones13:59
XarkYeah.  For things like this I think consistency with other archs trumps "aesthetic concerns".14:00
blueCmd woohoo my first relocations! \o/14:04
XarkNice.  Working on TLS (or just general linker stuff)?14:05
blueCmdTLS, I want to port eglibc to openrisc14:05
blueCmdand the recommendation from the glibc guys at #glibc was that I s14:06
blueCmdshould implement TLS first14:06
blueCmd(I have some compilation dependecy loops with emutls that GCC uses which I hope this will solve)14:06
XarkI see.  Makes sense, as  I can see it being a prerequisite.14:07
stekernblueCmd: nice progress!15:03
stekernblueCmd _franck_, I've added you to the "GNU toolchain" openrisc github team15:07
stekern_franck_: so once you are happy with your gdb work, feel free to push that15:08
blueCmdstekern: cool15:12
stekernnow if I only could get this immu stuff working...15:34
_franck_thanks stekern15:45
_franck_I just wanted to run the regression test on it15:45
_franck_however, I'm an expert and it's not my cup of tea...15:46
_franck_so if someone can try to run the testsuite I'd be more than happy :)15:47
blueCmd_franck_: ah, gdb - great! I will need gdb soon.17:13
blueCmdhow do you use gdb? native via simulator, qemu or something else?17:14
_franck_I'm using it with openocd server17:35
stekernblueCmd: for toolchain debugging, I recommend using or1ksim and it's tracing facility19:13
blueCmdstekern: I never got the tracing to work as I want19:32
blueCmdthe disasm says that the instructions it's executing are invalid and stuff like that19:33
blueCmdmaybe I have a faulty version of or1ksim19:33
stekernblueCmd: probably things that are still not fixed on 64-bit20:09
stekernI've got a 32-bit version I'm using20:09
stekernit should really use the opcode library though, but that hasn't happened yet20:10
stekernI think it has some stuff copied from bfd into it20:10
blueCmdanyway, my GCC is now generating TLS code, the assembler is creating reloc "targets" or whatnot for it and marking the symbols for TLS20:11
blueCmdthis is fun! I need to figure out why -O makes gcc forget about TLS though20:11

Generated by 2.15.2 by Marius Gedminas - find it at!