IRC logs for #openrisc Wednesday, 2014-01-29

--- Log opened Wed Jan 29 00:00:15 2014
blueCmdstekern: cool stuff! :D05:06
blueCmdlet me know if yoy need debian to it ;905:06
stekernblueCmd: =)05:57
blueCmdstekern: is it like 08 at your place now?05:58
stekern07:58, so around 08 yes05:58
blueCmdI was 'what on earth is he doing up at 0600 a Saturday?'05:59
blueCmdthen I remembered that 1) you're in finland05:59
blueCmdand 2) it's not saturday05:59
stekernand 3) stekern wakes up around 05:00 even if it's saturdays ;)06:00
blueCmdI think you're broken06:01
stekernyes, I suspect there is something wrong with the sleep function06:03
blueCmdyes, normally it's guaranteed to sleep at least the given amount of time06:05
stekernI think mine works by doing at most the given time06:10
olofk_stekern: Cool stuff06:41
olofk__franck_: Do you remember the details around the patch in adv_debug_sys07:14
jeremybennettblueCmd: stekern: You don't need lawyers. The official way is for all the contributors to have an assignment agreement with the FSF *and* for them then to agree to assign the work.07:22
jeremybennettThe alternative is for them to assign the work to someone who already has an FSF agreement, who can then assign it. We quite often do that for customers.07:23
jeremybennettNote that the person doing the assignment is saying *two* things. First that they wish to assign ownership of their work to the FSF and secondly (the important one) that they have the right to make that assignment.07:24
jeremybennettThe challenges for OpenRISC are first that some of the major contributors haven't been part of the project for a decade and not all are still friendly to OpenCores07:25
jeremybennettand second that some of them did the work while working for another company (Flextronics), who probably ultimately owns the work.07:26
jeremybennettThe first step is to go through and work out who is genuinely a contributor. You could for instance argue that I pretty much rewrote GDB from scratch, so the previous contributors no longer have any rights.07:27
jeremybennettBut you'd need to check the GDB 5.3 and current GDB code bases to be sure of that.07:27
jeremybennettWhere you have people you can't contact, or who for whatever reason are unable to contribute their code, you then have to rewrite it - ideally in a clean room, so there can be no claim of inadvertent copying.07:28
jeremybennettBTW - I have a FSF contribution agreement in place and I am very willing to contribute anything I wrote!07:29
jeremybennettThe FSF are generally pretty helpful with this sort of thing, so it might be worth an email to them. We won't be the first project in this situation!07:31
jeremybennettI don't think I made any significant changes to binutils. Joern Rennecke may have done when he was working on GCC 4.5.1, but he is pretty disciplined about ChangeLogs and so you'll find it in there (possibly as amylaar).07:35
jeremybennett(and since he did the work as Embecosm there will be no problem committing his work).07:35
jeremybennettThe problem you have are the mystery contributors between Johan Rydberg and around 2008, since there was quite a bit of change. Not everyone was so disciplined about ChangeLogs (sigh), so you'll need to slog through the SVN and CVS records07:37
jeremybennettBTW - I think Johan lives in Stockholm and juliusb knows him.07:37
stekernjeremybennett: yes, and that's why we want to start with the easy part, binutils. There are no problems with tracking down people there, there are no mystery contributers in binutils between Johan Rydberg and Juliusb 201107:41
stekernand everything that Johan Rydberg wrote is already upstream, so no issues with tracking him down07:44
olofk_Could someone generate diffs from binutils-changed? I don't have the debian tools available08:05
stekernisn't this the binutils diff:
_franck_web__olofk_: this is because when openocd first connect, registers content is undefined. In this case, the computed CRC is wrong and OpenOCD gives up....08:23
olofk_stekern: Ah.. didn't see that08:25
_franck_web_we could workaround this by adding a flag in OpenOCD to disable CRC check08:30
olofk_hmm.. I'm looking at the or1k-src repo on github now, and matching pgavin's inital commit against the or1k-debian diff09:52
olofk_The problem is that there is also a README-or1k file that looks like it was written by juliusb, so did juliusb or pgavin do these changes?09:52
olofk_ok... tracked it down to
olofk_I hope this wasn't copied from somewhere else :)10:02
stekernyes, it's a bit unfortunate that peter destroyed julius history when he imported it10:16
stekernI had planned to rebuild that on the rainy day10:17
stekernlet's see what happens first, or1k-binutils upstream or a rainy day10:17
olofk_Does anyone know an easy way to add comments to text files? I'm looking for something like the inline comments on github11:47
olofk_Trying to keep track of which changes that belong to a certain commit11:50
olofk_In the binutils diff11:50
olofk_ahh... a word processor works fine!11:53
olofk_I knew they were useful for something11:53
stekernolofk_: try git blame instead11:54
olofk_stekern: Yes, I'm using blame, but pgavin merged julius original work with his own changes in a single commit11:55
stekernso merge the two git blames, anything that's from the initial commit should be replaced with the blame from julius repo11:55
jeremybennettolofk_: stekern: You might want to think about migrating to the new binutils-gdb git repo as your baseline. IIRC pgavin's work used a home grown combined repo that simoncook and I put together.11:56
stekernjeremybennett: that's the rainy day I'm speaking about...11:56
stekernand he didn't use your repo, he just checked out the src directory from cvs11:56
olofk_jeremybennett: Yes, we should to that, but for now, I'm just trying to record the currrent deviations from upstream11:58
olofk_bfd/aclocal.m4 is autogenerated, right? So we shouldn't care about that?12:07
olofk_oh.. that wasn't touched at all in bluecmd's version. Never mind12:09
olofk_jts_: Welcome12:09
stekernolofk_: I think FSF has a rule for insignificant changes that is less than 20 lines or something like that12:10
stekernand pure boilerplate in config files and such are probably included in that as well12:11
olofk_So I shouldn't worry about and friends either then12:12
stekernI wouldn't12:13
olofk_That makes it a bit easier12:13
amsolofk_: yes, i have (quite alot) of experience with the fsf assignment stuf ...12:14
amsolofk_: but for it to go upstream, it will have to be GPL.12:14
amsolofk_: since we are talking about binutils; which is GPL ...12:15
amsolofk_: but ... the fsf grants back a royalty free license to do whatever you want ... so ..12:15
stekernams: you have to be less resistant to olofk_'s jokes ;)12:17
olofk_Could you at least consider to move to MS-PL. I think that's the future. Might be a good idea to rewrite it in C# at the same time12:17
olofk_winutils? :)12:18
amsstekern: ?12:18
amsolofk_: oh .. trolling ...12:18
amsi thought you were serious :-)12:18
olofk_Sorry about that12:18
amson a serious note ... anyone in paris here?12:19
-!- jts_ is now known as jtdesousa12:20
amsi need housing ... preferbly for 1-2 months ..12:20
stekernI think our closest french connection is _franck_web_12:21
stekernand ysionneau12:21
amsbut 2-3 weeks will do .. i'm currently looking for places to stay, but fuck it is hard ...12:21
olofk_hmm.. how can I do a blame on a deleted line? Looking for this change in bfd/archures.c "-extern const bfd_arch_info_type bfd_openrisc_arch; "12:25
amsolofk_: changelog?12:26
olofk_ams: I'm more or less trying to reconstruct the ChangeLog from two different repos12:26
ams... why?12:26
amsbetter not i ask ..12:27
stekernolofk_: why are you trying to track deleted lines?12:27
olofk_I found it, so it doesn't matter12:28
olofk_I guess that a deleted line would count as a change too, right?12:28
stekernI don't think anyone can sue you for deleting copyrighted material ;)12:28
olofk_hmm... true :)12:28
stekernIANAL though... anything is possible12:28
amsI can give you one legal advice that no lawyer would disagree with ...12:30
ams"It depends."12:30
stekernhaha, that's good advice12:30
amsOh, and I do have another legal advice to...12:32
ams"Talk to a lawyer."12:32
_franck_web_ams: try ysionneau I think he is living in Paris area. I'm living in France but I'm close to Nice (south east)12:32
jonibodeleted lines won't be submitted upstream so no need to know who wrote/deleted them12:32
ams_franck_web_: thanks!12:33
olofk_jonibo: Why not?12:33
amsjonibo: ah, but the devil is law, what if you only renamed the function?12:33
jonibothen the renamed function name exists so there's a line attributed to that person...12:33
amsjonibo: what if the naming of the function helped in creating another better version, i.e. it being a deriviate of the original ...12:33
jonibobut that's a trivial path where the FSF won't care anyway12:33
amsjonibo: uhm, no, they will care.12:34
olofk_To give some context, I have found a few of these:12:34
olofk_-extern const bfd_arch_info_type bfd_openrisc_arch; xtern const bfd_arch_info_type bfd_or32_arch; +extern const bfd_arch_info_type bfd_or1k_arch;12:34
olofk_oops, no line breaks12:34
joniboand you want to know who wrote the original deleted lines?12:34
olofk_jonibo: No, but to be picky, juliusb did the two last lines (delete+add) and pgavin the first one (delete)12:35
olofk_So, one could say that juliusb did a rename, and pgavin a delete12:35
jonibobut only the added line gets submitted upstream12:35
stekernexcept if the deleted lines are already upstream12:36
olofk_jonibo: Why? Shouldn't we send the deletions upstream?12:36
joniboare they?12:36
olofk_What I pasted was a diff from upstream to our port12:36
jonibono, just munge everything into one big change and submit it as openrisc support12:36
amsjonibo: it doesn't matter if it is the only one getting submited upstream; a delete changes the source code in a deirivate manner, so that can also be important12:36
jonibocopyright assigned to the FSF12:36
stekernbut the diff isn't against that file anyway, since we've renamed the whole from openrisc to or1k12:37
jonibobut the change is already upstream so is owned by FSF12:37
jonibothe deleted lines are already upstream12:37
jonibowhich project... binutils, right?12:38
joniboso every line that's already upstream is already owned by FSF12:38
jonibothe openrisc file, even when renamed to or1k, is still owned by them12:38
joniboand whoever makes changes against that file owns the change and needs to assign copyright to FSF12:39
jtdesousa /msg nickserv jtdesousa info12:39
joniboams: just to comment on what you said above... changing the code in a derivate manner just means that the derived works needs to be GPL... the copyright of the changed code is owned by the new author.  The FSF only cares that the copyright can be transferred to them, and that can be done exclusively by the new author... it doesn't matter who wrote the the code that "inspired" the new work if none of the inspiration remains after the rewrite12:44
_franck_web_is there some senior embedded developers here ? I'm working on a VME system running OS9....I wasn't born was this board was designed :)12:45
amsjonibo: that is completely bogus... sorry.12:47
joniboin what  way?12:47
amsjonibo: 1)the changes do not have to be GPL.12:47
amsjonibo: 2) the FSF cares about the whole copyright chain, who wrote it, who helped debugging it, if you had a coworker sitting beside you then it also matter12:48
amsjonibo: 3) if the work was rewritten, then the original code still matters, specially if one was inspried by it... you based it on a differnet work, even if you rewrote it.  (depending on the type of rewrite)12:49
jeremybennettjonibo: ams: It helps to break the port up into useful structures - submit it as a group of patches, each dealing with one aspect. Then you are more likely to get it reviewed.12:49
jonibowith that logic, I could add a comment to the code and then remove the comment and then refuse to assign copyright for my non-existent code to the FSF and you'd be screwed over forever because I could always claim that my code (which isn't there) is part of your copyright chanin12:49
amsjonibo: incorrect logic, your comment is not a deriviate work, a rewrite touches many more aspects than a small block.12:49
jeremybennettYou don't have to worry about completely trivial changes. There is an exception for that, and a single line of comment is almost certainly considered trivial.12:50
joniboams: why don't the changes need to be GPL?12:51
amsjonibo: they have to be a compatible license.12:51
jonibopoint 2)... the FSF cares only who owns the copyright... if my employer owns it, then it's moot who was sitting beside me when it was written12:51
amsinetutils for example is both modified BSD and GPL; but the whole work as such is GPL; but the original code from BSD is still licensed under the original license.12:51
amsjonibo: they care about alot more than that12:51
jonibooh, fine, i'm not interested in splitting hairs though...12:52
joniboBSD or GPL then12:52
amsjonibo: or LGPL, or ...12:52
amsMIT, X11, ...12:52
amsany compatible license.12:52
jonibowhat does this have to do with inetutils???12:53
joniboi thought we were talking about bnutils12:54
amsjonibo: an example where "don't the changes need to be GPL" does not apply12:54
amseven binutils has some BSD licensed code12:54
joniboso you're preaching to me about licensing???  way off topic, sorry12:54
amsjonibo: it is clearly on topic, since you are saying things that are clearly false with regard to licensing.12:54
jonibowith respect to binutils?12:55
amsjonibo: and the topic _was_ licensing.12:55
stekernyes, olofk_ wants to submit it as MS-PL!12:55
joniboanyway, code ownership matters... if you borrow code from someone else, they own it... if you rewrite it based on their example, the licensing restrictions apply, but ownership does not apply... FSF only cares about ownership12:56
amsjonibo: sighs, that is still wrong, so wrong.12:56
joniboyou'd better look up some other examples and you'll see i'm right12:57
amsjonibo: uhm, ok, so i take openrisc, look at the code, and then write the same thing, then my work does not fall under the openrisc copyright?12:57
amsjonibo: and i work of the code...12:57
jonibowhich openrisc copyright?12:58
amsjonibo: ok, i'll put this to rest, you are wrong, terribly wrong.  a rewrite does matter if you base the rewrite on someone else work, a rewrite is a deriviate work, period, the license does not matter in that case.12:58
amsif i take some work by you, look at it, and think i can do better, i am basing my work on yours, then it is a derivcate work.12:59
joniborewrite is a derivative work... yes, absolutely... so license applies12:59
amsend of story, over and out from me.12:59
ams(now to hunt bugs)12:59
jonibointeresting cop out... can't be bothered to explain why you're right so you just walk away... I guess I'll have to continue being "wrong" then since you won't enlighten me... sigh13:00
stekernif that would be true, then every pop song made would be sueable (spelling?) for copyright infingement13:00
olofk_Who deleted bfd/coff-or32.c? git gurus, bring it on!13:00
amsstekern: most are ...13:00
amsstekern: problem is finding the "original" copyright holder13:01
olofk_Because unless we all agree how deleted stuff should be handled, I will at least write it down to have a complete picture13:01
amsolofk_: do that13:01
stekernyeah, that seems like less trouble than to figure out if it matters...13:02
joniboif you don't delete FSF code, it's already owned by the FSF so you don't need to worry about it... if it's somebody else's code, then it might make sense to do so13:02
joniboi'd just make that disctnction to save yourself some effort13:02
jonibobut do what you'd like... it's your time :)13:02
joniboglad to see you undertaking this, in any case... it's been a long time coming13:03
stekernanyone that has deleted code have probably added a bunch of code that hasn't been deleted anyway13:04
joniboexactly... just focus on who has added code and assume they are everybody13:05
joniboif somebody obvious is missing, then do something about it... otherwise it's certainly fine13:05
jonibowhat's the worst that can happen, anyway13:05
jonibo... code must be retracted and a couple of small changes need to be made for trying again?13:05
joniboas far as I understand, binutils is a pretty clean rewrite by juliusb et al. anyway... we know who the authors are13:06
olofk_Almost done. I'm on page 4 of 462 of the diff now :)13:06
joniboawesome... did pages 6 and 2 take you long?13:07
olofk_In case you are wondering, this is what I'm doing. Will that be ok? Takes a little time, but I can do a little now and then
olofk_15/462 done13:18
olofk_four authors found so far (pgavin, juliusb, stekern and bluecmd)13:19
joniboi think that's easy enough for deletions... which I don't think you need to worry about anyway13:27
jonibogit blame will give you the "author" of the added lines13:27
jonibothe tricky part is when the "author" copied the code from the old implementation (which it looks like there's a fair amount of) and you need to figure out who wrote that13:28
jonibodoesn't opencores just own all this code based on the fact that they claim copyright on anything you push to their server?13:29
jonibo...would be so much easier to beat everybody down with that stick13:30
olofk_jonibo: (Ignoring the bitterness) This stuff was all done after we moved it to github13:34
joniboyeah, I know... I was being facetious and hoping for an easy way out13:34
olofk_...I hope at least... fuck... didn't consider that we might need to check if juliusb copied it from SVN13:34
joniboI think he copied large swathes13:34
jonibothe relocation, code, for example, looks familiar13:34
jonibothen again, all relocation code for all arches looks kind of the same13:35
olofk_yes, I guess some things here can be considerered trivial, which might help if it becomes unclear who was the original author13:36
jonibobut, honestly, I don't think that you need to be this thorough... we know who the 10 people are who did "major" work on this... and the odds of a minor player showing up and complaining that you upstreamed their work is just too small to bother worrying about13:36
jonibojust get a sign off from the 10 people who matter and push upstream13:37
olofk_I'm not worried about the risks, just that FSF could be anal about this, so I think it's easier to just do this once and for all13:37
jonibonobody expects a hobby code to become a hobby lawyer... just play dumb if somebody actually questions you13:37
jonibo"we thought that was everybody"13:37
jonibogood enough13:37
jonibootherwise we won't get any orpsocv3 improvements during 201413:38
olofk_jonibo: hahaha. You know how to hit me where it hurts :)13:38
joniboit's kind of like patent infringment... the less you know, the better off you are... "oops, sorry about that"13:38
joniborectify whatever mistakes you make after the fact... I'd be willing to bet money that you get no pushback from anyway... FSF or otherwise13:39
jonibojust as long as the major players 99.9% of the code are on board13:39
jonibogood enough13:39
joniboit's a hobby project, after all13:39
jonibo(for most of us)13:39
joniboand if this ever becomes wildly successful and you're filthy rich, you can pay off whoever comes complaining that you robbed them of their "glory"13:40
jonibobut, like I said, it's your time...13:41
joniboand I still (really) appreciate the effort to get this upstream13:41
jtdesousaolof_k: hi, jose here. Interesting conversation ...13:44
jtdesousajonibo: good points13:44
_franck_web_this is why I contribute to orpsocv3, in case olofk_ becomes very rich :)13:45
_franck_web_hi jtdesousa13:45
jtdesousahi _franck_web_, me too13:46
olofk_jtdesousa: Hi Jose, you caught us in the middle of one of these never ending license discussions :)13:46
jtdesousaolof_k: sorry my ignorance, any of you with opencores?13:47
olofk_jtdesousa: I used to work for ORSoC who owns OpenCores13:48
olofk_and LoneTech works there now13:48
jtdesousaolof_k: ok thanks. IMHO, everything should be on github...13:49
olofk_jtdesousa: I both agree and disagree. :)13:50
olofk_Got to run. Be back later13:50
jtdesousaolof_k: need to understand that better13:50
olofk_How is binutils structured internally? Would it make sense to start with sending in the changes for everything under bfd first?15:06
stekernthe relocations are based off the openrisc port, so they actually have quite different names from the or32 one (SVN). that said, that code is as much boilerplate as it can be and looks just about the same on every arch15:08
jeremybennettolofk: It certainly makes sense to keep bfd, binutils, gas and ld patches separate (as part of a group though). Then experts in each can review.15:12
blueCmdams: I have french co-workers, I could ping them if they have something to rent15:23
olofk_ok, then I'm done with bfd. I scanned through the SVN logs too and found additional changes by Joern15:23
amsblueCmd: please do!15:24
amsblueCmd: anything short term, and anything long term is of immense interest .. specially short term15:25
olofk_Jesus... I hadn't looked at the SVN repo for a while. We store quite a few toolchains down there :)15:25
olofk_jeremybennett: oh, but you mean that we still should send all the patches at once?15:27
stekernolofk_: yes, that's right, now I remember, I have pulled in two(?) bugfixes by Joern. I was careful to add his changelog entries to the or1k one though.15:30
blueCmdstekern: I only see Joern in GCC15:40
blueCmdan he has an embecosm email so that should be fine15:40
blueCmdolofk_: excellent work btw!15:43
blueCmdso I suggest doing this jonibo's way that we just do:15:46
blueCmdgrep '^[0-9].*' or1k-src-changelog-combined.or1k | cut -f 3- -d ' ' | cut -f 1 -d '<' | sort -u15:46
blueCmdwhich gives me, _franck_, pagvin and stekern15:46
blueCmdhm, why doesn't juliusb show up?15:47
blueCmdanyway, get the ball moving and see if they really need us to give them commit by commit15:48
jtdesousaare you guys planning an official release of the toolchain aspart of Gcc?17:57
stekernjonibo: do I remember correctly that you were saying that you've been seeing arg regsclobbered by syscalls?19:02
-!- jjts is now known as jtdesousa19:56
_franck_jtdesousa: that's the plan. However, this not easy so the first step would be binutils20:12
-!- jjts is now known as jtdesousa22:46
--- Log closed Thu Jan 30 00:00:16 2014

Generated by 2.15.2 by Marius Gedminas - find it at!