| lkcl_ | it's also a good one because this is heavily-tested code (the decoder) containing literally tens of thousands of entries | 00:00 |
|---|---|---|
| lxo | and it would lead to the problem he brought up if we were to go back and modify the vhdl file and rebuild the cvs file from it and reapply the manual updates | 00:00 |
| lkcl_ | the absolute last thing you want is to replicate that information by hand | 00:00 |
| lkcl_ | "just because of copyright assignment" | 00:00 |
| lxo | no, the issue is what the source code is | 00:00 |
| lkcl_ | that i would say is a good thing | 00:01 |
| lxo | when there's editing on both forms, neither is clearly the source | 00:01 |
| lkcl_ | by having a verif... | 00:01 |
| * lkcl_ thinks | 00:01 | |
| lkcl_ | i was going to say: you want an auto-generator and you *definitely* do not want to be editing it | 00:01 |
| lkcl_ | but if you did - by mistake - you want a program that can tell you | 00:02 |
| lxo | there's an issue of freedom (without sources you're toast, or at least at an artificial disadvantage), and there's an issue of compliance with the gpl | 00:02 |
| lkcl_ | "this is what the output *should* be like, but someone's clearly edited it" | 00:02 |
| lxo | but for binutils, there is also the issue of copyright assignment | 00:02 |
| lkcl_ | this is tricky | 00:02 |
| lxo | because these are really distinct issues | 00:02 |
| lkcl_ | then a discrepancy-checker would be essential for that | 00:02 |
| lkcl_ | yes | 00:02 |
| lxo | we have a process to make it clearly source | 00:02 |
| lkcl_ | if this was 5 lines of code it would not be burdensome to go "screw it, re-type it" | 00:03 |
| lxo | but we don't have a process to detach it from others' copyrights | 00:03 |
| lkcl_ | however this is *tens of thousands* of fields | 00:03 |
| lxo | now, there's another angle to look at: facts are not copyrightable | 00:03 |
| lkcl_ | burdening someone with replicating that by hand is not safe. | 00:04 |
| lkcl_ | ahh... interested to hear where this one goes | 00:04 |
| lxo | so extracting the facts | 00:04 |
| lxo | into a newly generated file | 00:04 |
| lxo | could keep us in the clear | 00:04 |
| lxo | we may have to talk to a lawyer about this | 00:05 |
| lkcl_ | there's another aspect: this information originally came from the IBM Copyrighted v3.0B Technical Reference Manual | 00:05 |
| lxo | I'm sorry I didn't notice this potential pitfall before | 00:06 |
| lkcl_ | yeah i didn't think of it either | 00:06 |
| lxo | I suppose it's not freely-licensed | 00:06 |
| lkcl_ | can i leave it with you, some links, feel free to cc me on conversations with richard if you like | 00:06 |
| lkcl_ | https://github.com/antonblanchard/microwatt/blob/master/decode1.vhdl#L196 | 00:06 |
| lkcl_ | IBM's default license for source code from all researchers is CC 4.0 | 00:07 |
| lxo | ibm does have a copyright assignment on file with the fsf, but I suppose it would only cover stuff they actively contribute to a covered project | 00:07 |
| lkcl_ | here's the one where that information is "extracted" | 00:08 |
| lkcl_ | https://git.libre-soc.org/?p=libreriscv.git;a=blob;f=openpower/isatables/minor_31.csv;hb=HEAD | 00:08 |
| lkcl_ | actually... we could ask! | 00:08 |
| lxo | yes, indeed | 00:08 |
| lxo | as for how to name the branch... I went for a per-user WIP branch mainly because I didn't want to choose the name for our development branch in the global namespace, and because of the issue I'd raised before, about non-fast-forward pushes, that is often found tolerable in WIP branches, but not in branches for public consumption | 00:56 |
| lxo | I learned that naming branches can be a very sensitive and controversial matter. just the other day, I read a social media post that went "in order to sidestep the master vs main controversy, I've named the development branch of my project 'donald-trump'" :-D | 01:54 |
| programmerjake | that reminds me of the discussions about what to name the error returning operator in Rust -- since Rust uses enums for non-fatal error propagation and just returns them from a function by value (instead of using stack unwinding like C++), | 03:33 |
| programmerjake | they've been arguing about if they should use raise, throw, fail, or some other keyword, so, they settled on using yeet as an agreed-bad-by-everyone placeholder till they decide which keyword they want to use | 03:35 |
| lxo | yeet?!? I don't know that word. does it even mean anything (other than the rust-assigned meaning)? | 04:49 |
| programmerjake | yeet is a newish (slang?) word where there isn't really an agreed definition, but it generally means "to throw something with excitement" or is an exclamation. those meanings are unrelated to rust and/or programming in general | 05:36 |
| programmerjake | lxo ^ | 05:37 |
| jn__ | on branch naming: in gitolite it's possible to give everyone a private prefix (e.g. lxo/*, jn/*, etc.), so that each person can still have any number of branches, but they don't collide | 08:45 |
| jn__ | ok, when people call the main branch donald-trump, that's just controversy for controversy's sake | 08:46 |
| lkcl_ | lxo: Alain Modra kindly pointed out a couple of things: Anton Blanchard is already on the IBM->FSF Copyright Assignment list | 10:31 |
| lkcl_ | and that the CSV files constitute a "specification" | 10:33 |
| lkcl_ | which has been "extracted" | 10:33 |
| lkcl_ | the other thing is: argh, actually, sv_analysis.py *used* to be just one file but the reading and processing of CSV files to create SVP64 CSV files requires several other modules | 10:35 |
| lkcl_ | lxo: i have an idea. extract everything related to OpenPOWER v3.0B for the ISACaller simulator and make it a GNU project | 10:58 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!