IRC logs for #openrisc Thursday, 2015-01-22

--- Log opened Thu Jan 22 00:00:07 2015
-!- heroux_ is now known as heroux02:06
-!- clopez_ is now known as clopez02:16
jeremybennettolofk: Closing or32 bugs is fine by me.  We're putting all the new work into the or1k tool chain08:50
poke53282git really confuses me.08:51
stekernpoke53282: you mean you don't 'git it'?08:53
stekern(pun stolen from John Romero -
poke53282In the main qemu repository I can find parts of my code, that to my knowledge were not accepted by Jia Liu ( maintainer of the openrisc part).08:54
poke53282stekern: Exactly.08:54
poke53282Ok, this is fine08:54
poke53282But I can't find in the history of the file an entry the last year, who put it there.08:55
poke53282I loocked up my patches and the commit patches. I wrote that I included some variables. And I can find them in the main repository.08:56
stekerngit blame?08:56
poke53282How do I search for the first patch, which introduced a variable?08:58
stekernhave they changed since they were introduced?09:01
stekernyou can use git log -L09:02
stekernto show a subset of lines09:03
olofkjeremybennett: Good to know. Do you have any input on the remaining open bugs? Especially 21, 66 and 10309:06
olofkFuck it. I'm closing 66. It works for or1k09:06
stekernpoke53282: appearantly you can use -S'string' to search for something to09:10
stekerni.e.: git log -S'variablename' -- path/to/file09:10
jeremybennettolofk: Sounds a good idea09:11
jeremybennett21 should be transferred to or1k - it is an issue that needs resolving09:13
jeremybennett103 needs checking that newlib is correctly configured in the standard scripts for or1k, then can be closed09:14
olofkjeremybennett: Regarding 103, what is "correctly configured"? Are you saying that we should enable long long by default? I don't have much insight in how widely used this feature is or how much extra code it will bring in09:19
olofkSome arches set it as default. Other don't09:22
olofkwallento: What's the status on upstream newlib now? Did a fresh install (using upstream newlib git), and or1k-elf-gcc can't find stdio.h09:57
wallentoonly the latest patch is missing (mlock)09:57
wallentowhen I apply this patch to upstream and build, it works fine09:58
jeremybennettolofk: I think we should support long long - it's part of C99.  I didn't realize that newlib did not by default.10:15
poke53282git rebase drives me crazy10:17
stekernif it drives you crazy,  don't use it10:18
stekernmerge instead10:18
poke53282If there is a conflict, I should solve it and then do a "git add"10:19
poke53282and then git rebase --continue10:19
poke53282but it doesn't look like, that the patch is changed afterwards.10:19
stekernbut did it change?10:20
poke53282I did a rebase one time last week and fixed conflicts.10:20
poke53282And now it seems, that I have the same conflicts again.10:20
poke53282or maybe others. Not sure.10:21
poke53282stekern: I would like to have my changes on top and not in between.10:22
poke53282If I want to send this patches to a mailing list, I have to do something like this anyhow.10:24
poke53282I think, that it is much better, to do it now, and not when whole stuff gets even more complicated.10:26
poke53282My problem is, that my second rebase doesn't work as expected. If I look at the patches with "git format-patches", it looks like, that they are unchanged10:40
poke53282Is there a changelog for a commit?10:43
ysangkokgit log <commitid> ?10:44
poke53282With git rebase I change an already commited patch.10:46
poke53282for example the title.10:47
poke53282But I don't see an information like "Changed at Dec. .... 2014" or something10:47
ysangkoki think it would be best to ask in #git :P11:01
olofkjeremybennett: That sounds like a good enough reason to support it. I'll cook up a patch and close the bug when that is applied11:05
stekernpoke53282: afaik, there's no history11:19
stekernyou can probably dig the old commit out from the reflog though11:20
olofkAha. I didn't know about git rebase. Only used git pull --rebase11:47
poke53282well, I have my commits in another branch.11:51
poke53282So I pull master first, and then I do a "git rebase master" on my branch11:51
poke53282Let's try an interactive rebase and change "pick" to "edit"11:53
poke53282stekern: Yes, the local reflog contains such information in some way.11:54
poke53282I simply want to know, why I can't find my previous solved conflicts.11:55
poke53282and why the commits are not updated with these solved conflicts.11:56
olofkwallento: Just to check. Did you compile anything with the rebuilt tc?12:09
olofkI could also try doing it again with your upstream branch instead of master :)12:10
olofkwallento: I'm confused too now. Which repo and branch should I use?12:12
olofkHmm.. does git branch <name> automatically create a new branch <name>?12:14
olofkBuilding for-upstream from open12:16
olofkrisc/newlib now12:16
olofkHope that's correct12:16
wallentoyes, thats correct12:23
wallentothats upstream plus a fix12:23
poke53282After, I have run rebase three times, everytime with other conflicts,  it doesn't complain anymore.12:23
wallentogit branch <..> does not work12:23
wallentoeh, solve the issue12:23
wallentogit checkout for-upstream creates new branch and goes to it12:24
wallentobuild works but not for upstream :)12:24
poke53282It gets even better. He removed some of my commits.12:41
olofkwallento: Not following you now. Have you seen problems with your for-upstream branch?12:55
olofkpoke53282: Sounds like you're not entirely impressed by the state of qemu right now :)12:59
poke53282well, I am not impressed by the way git handles my rebasing.13:00
poke53282But I think, it is done.13:01
poke53282QEMU doesn't run my openrisc binaries anymore. But otherwise I am fine.13:03
wallentoolofk: nope, I actually built from it13:05
wallentoand use it also13:05
wallentofor my software13:05
wallentotogether with binutils upstream13:05
olofkI suck :(13:11
ysionneau1:48 < poke53282> But I don't see an information like "Changed at Dec. .... 2014" or something  < if you change the commit , the sha-1 hash will change, right?13:25
poke53282ysionneau: Yes, looks like this is the case13:30
ysionneaudon't know if that was your issue when you said you fixed merge issues while rebasing but then you didn't see those changes when doing git show13:36
poke53282I have reset everything and have redone the rebase. It seems to work now.13:38
ysionneauok :o git mistery then13:42
wallentoolofk: still can't build? whats the error?13:46
olofkIt's building but can't find stdio.h when I compile a hello world program13:51
olofkwallento: "echo | or1k-elf-gcc -v -x c -E - " gives me some warnings14:12
olofkignoring nonexistent directory "/opt/or1k-elf/lib/gcc/or1k-elf/4.9.1/../../../../or1k-elf/sys-include"14:12
olofkignoring nonexistent directory "/opt/or1k-elf/lib/gcc/or1k-elf/4.9.1/../../../../or1k-elf/include"14:12
olofkThat can't be good14:12
olofkotoh it finds  /opt/or1k-elf/lib/gcc/or1k-elf/4.9.1/include14:14
olofkAnd there are a bunch of stdarg, stddef, stdint and stuff there, but no stdio14:14
olofkBut who uses stdio.h nowadays. I use a docker app running in the cloud instead14:15
wallentommh, works fine for me14:37
wallentoso, lets check if everything is identical14:37
olofkOk, let's do this carefully. What's on the first line of your
olofkAfter we are done with that file we only have 5955 more files to compare14:39
wallentobinutils is at 6c1965f995bc8119c0775f2f9b648a1902e6e21014:40
olofkI'm using the 2.25 tarball14:40
wallentomine should be the binutils-2_25 tag14:41
wallentoor1k-gcc is 50e11f73098c722676829ae73caf09246e81cd44 correct?14:41
olofknope. Mine is probably one commit behind yours. Is the previous commit 1ee8ef4190f85934479fa00abbc29e164af80f00?14:43
olofkDid a pull now and got the same as yours, just in case14:43
wallentookay, sorry, my boss needs computer support, will be back in an hour ;)14:44
olofkJust a quick one. newlib is 5d9075b68b6d53ecd1895b77ee01ffb30a521bbf14:45
wallentoi will also run it again, just to be sure14:52
wallentoolofk: I am really sorry, but works for me15:14
olofkI'm more sorry :)15:34
olofkThe mlock patch was applied now btw.15:35
olofkDoes that mean that everything is upstream now?15:35
wallentoyet another beertime15:35
olofkAhhh... so that's why :)15:35
wallentogetting upstream is a path ;)15:36
wallentoso, this one is "normal": ignoring nonexistent directory "/opt/or1k-elf/lib/gcc/or1k-elf/4.9.1/../../../../or1k-elf/sys-include"15:37
wallentoat least I have the same15:37
wallentothe other one disappeared after installing newlib15:37
wallentodid you forget to "make install" ;)15:37
olofkI think it15:39
olofkAha! make install fails now15:39
olofkor1k-elf-ar: ../sys/lib.a: No such file or directory15:40
olofkwait a minute...15:40
olofkoh... wrong newlib15:41
olofkOh well. I'll wait until upstream git is updated and try again after that15:41
wallentosounds good15:53
wallentoolofk: GO!16:22
wallentoits in the git16:22
poke53282blueCmd: It is Ok, if I cherry pick some of your patches for QEMU?16:52
poke53282cherrz pick16:53
blueCmdpoke53282: of course16:54
poke53282I might merge and move them too. With my skill in git you might accidently lose official authorship ;)16:58
blueCmdpoke53282: yeah, that's fine. I'm sure you will do something sensible17:00
-!- _franck__ is now known as _franck_19:19
reinoudhi folks20:37
reinoudis this also the place to talk about riscv ?20:37
ysangkoknice troll :D21:29
poke53282But indeed, looks like the riscv people don't have their own irc channel yet.21:31
poke53282there are a few persons in #riscv21:32
poke53282reinoud: I would suggest, that you ask this question on the risc-V mailing list.21:33
poke53282This chat room is not directly related to risc-v.21:36
poke53282However, we have discussed RISC-V in general here. I don't think, that you can ask special technical questions here.21:39
poke53282bluecmd: l.lwa/l.swa support (always succeeds, no atomic protection) -> Semi-fake atomics, now with less fake ->  Eanbled stricter atomics, it still doesn't work for glibc but it's the Right Thing To Do -> Fix atomics22:37
poke53282Nice commit history22:37
poke53282Ok, I have cherry picked everything important. I will probably merge these patches to one.22:38
olofkwallento: Building now22:43
olofkI want to do an FFT on OpenRISC. Any ideas where to start?22:53
olofkwallento: Bring out the beer! It's working now with binutils 2.25 tar ball and upstream newlib git23:03
--- Log closed Fri Jan 23 00:00:09 2015

Generated by 2.15.2 by Marius Gedminas - find it at!