IRC logs for #openrisc Friday, 2016-04-29

--- Log opened Fri Apr 29 00:00:43 2016
shorneman this autoconf setup for gdb is hacky03:56
shorneok, I added another hack , but should be consistent04:09
wallentoshorne: you can also update the travis build04:10
wallentooh, well, you wrote04:11
wallentoshould have read all :)04:11
shorneyeah, but I am thinking this is just wrong for it to try to link in or1ksim when you have --disable-sim04:11
wallentowell, right04:11
wallentobut it did not try before, right?04:11
shorneit did before too04:11
shornewe just didnt have travis to point it out :)04:12
shorneI mean when it was in or1k-src04:12
shorneah, I made the fix now need to do autoreconf04:12
shornebut gdb says " error: Please use exactly Autoconf 2.64 instead of 2.69."04:12
shorneok, works now with disable-sim only04:21
shornenow with enable-sim04:22
shornefails to find or1ksim in default path (expected)04:23
shornewith using with-or1ksim=my path... it works04:23
shorneso its good04:23
shornepushing patch04:23
shorneideally we would disable or1ksim if the lib doesn exist using the AC_CHECK_LIB or AC_TRY... tests04:26
shornewallento: also, this way we dont have to update the build tutorial04:32
shornebut I guess we might want to since we "can" enable sim now :04:32
shorneif the code is merged :)04:32
shorneHope someone can review, the way I enabled nrun was a bit hacky storing the or1ksim_state in a random place04:33
shorneFYI, I havent tried the cgen (or1k) simulator, just the or1ksim wrapper stuff04:38
shorneok, travis build works too04:40
SMDwrkwallento: Hi, check email plz04:49
-!- Netsplit *.net <-> *.split quits: _florent_, rokka, kaalikahn, poke5328110:33
-!- Netsplit over, joins: rokka, poke5328110:37
wallentoSMDwrk: sorry, was really busy the last days12:58
wallentoI am answering now, is it okay to answer to the mailing list?12:58
wallentoshorne: merged13:16
wallentonext stop: upstream13:16
SMDwrkwallento: sure, no problem)13:54
olofkshorne: How is this sim thing working? IIRC we have the option to either use the built-in sim stuff from binutils/gdb or hook in or1ksim instead. I guess upstream don't want to depend on or1ksim15:08
olofkBut I'm not sure I have understood it all. ( jeremybennett ?)15:08
shorneolofk: you are right there are both the internal stuff (cgen) and the or1ksim wrapper.  The one I was testing is or1ksim16:32
shorneboth are merged16:33
shornewe can remove the or1ksim stuff, thats the one that has the kind of hacky autoconf setup16:33
shorneolofk: what you are saying makes sense, I have been testing with or1ksim, lets try to concentrate on or1k16:37
shorneThe build for sim/or1k is failing just tested16:38
olofkI think the problem for the internal sim thingie is that it requires some simulated peripherals to be of any use, and that might be a bit more work. Usual caveat, of course is that I don't really understand those parts of the toolchain16:49
olofkSo I'm kind of hoping that we can add gdb with adding sim stuff if it turns out to be a lot of extra work16:50
shorneI see, it seems peter gavin (from commit logs) was the one who wrote the internal simulator stuff. Do you know how far along it got?16:51
olofkDon't know unfortunately. And I'm not sure if Peter did all the work or if he is listed just because he did the initial github import and patch collecting. I vaguely remember that at least juliusb was digging into that as well16:56
olofkI'll throw Peter a mail and see if he remembers anything of it16:56
olofkdone. I'll let you know if I get a reply17:03
shornethere seems to be some changes in the way ENDIAN ness is declaird17:03
shorneneeding to fix that17:03
shorneok, now cgen (sim/or1k) is building17:04
olofkah cool17:04
shorneill run the testssuite with it17:04
olofkVery curious to hear the results17:05
shorne... Undefined target command: "sim".  Try "help target".17:05
shornethats not good17:05
olofkDo you need to rebuild gdb with sim support perhaps? Not sure how things are connected17:05
shorneyeah. thats what I thgouth it did17:06
shorneoh.. something was wrong with my config command the second time17:06
shorneok, more build failures to work through17:07
olofkwallento: I'm about to do a 180 on the vlnv syntax in .core files. I now think we should only allow a single v:l:n:v instead of separate v, l, n, v tags17:36
olofkThe reason for that is that we can then use the same syntax in the depend list, add a new VlnvList type in and parse into a vlnv object directly. No need to handle string parsing anywhere else17:38
shorneneed to regen the cgen files, which I figured out how to do, just need to pass --enable-cgen-maint to config18:13
shornebut now there is an issue, it seems the cgen guile (scheme) code does work with guile 218:14
shorneso the cgen fails, cgen was last released in 2009 :(18:15
shornemight have to go ask gdb developers18:15
shorneneed to go now18:15
shorneI was able to get it working with guile1.819:00
shornenow a failure in cpu/or1k.cpu, which is ok, making progress19:01
shorneok, finding some definite issues in cgen, checking with #gdb23:09
--- Log closed Sat Apr 30 00:00:44 2016

Generated by 2.15.2 by Marius Gedminas - find it at!