BitBake Apache License and Packaging for OpenBMC Components
What is the recommended way to handle license files for OpenBMC components?
I presume that any new code committed to the repositories should be Apache 2.0 unless special permission is granted. Based on a conversation regarding creating a StackExchange forum for OpenBMC, I presume MIT/OpenBSD-style licenses are verboten without special permission.
(1) What should the license tag be? (LICENSE = “Apache 2.0”?)
(2) Should a LICENSE file be created in the root directory? A COPYING file? Should we point to a common license file? With a checksum? If so, which one?
(3) Are there particular requirements or suggestions regarding how to package stuff?
(4) I presume SRC_URI file:/// links are okay since the git links are unknown and broken until the first version is checked in.
(5) How should I handle being in a subdirectory of a repository? The code I’m checking in will go into openbmc-tools. How does that work for CI?
I’ll go scan through the docs, but I didn’t find much commentary on these topics with light skimming. Just trying to get it right the first time.
Thanks in advance,
Wilfred
Quoted 4 lines
(5) How should I handle being in a subdirectory of a repository? The code I’m checking in will go into openbmc-tools. How does that work for CI?
There isn't any CI for openbmc-tools at the moment, as it was originally intended as a collection of quick hacks that made people's lives easier without any particular guarantees. This was done to lower the bar for entry and get people contributing their scripts.
CI is probably something we need to think about though, so I've added Andrew G in To/Cc:
Cheers,
Andrew
Quoted 6 lines · Andrew Jeffery
Quoted 5 lines
(5) How should I handle being in a subdirectory of a repository? The code I’m checking in will go into openbmc-tools. How does that work for CI?
Quoted 7 lines
There isn't any CI for openbmc-tools at the moment, as it was originally intended as a collection of quick hacks that made people's lives easier without any particular guarantees. This was done to lower the bar for entry and get people contributing their scripts.
CI is probably something we need to think about though, so I've added Andrew G in To/Cc:
I know for openbmc-build-scripts I did eventually make a jenkins job that does basic validation of the scripts it can (building docker containers). We were getting some regressions which were breaking our CI so this was a quick way to make sure the docker containers at least built when the scripts were updated.
We could do something similar with this repo, even if it's just running --help of the tools to ensure they execute. We could get more complex and connect them up to a QEMU session running openbmc to validate more. I don't have any bandwidth to do much here though so someone else would need to drive it.
Would we want a run-ci.sh in each subdirectory where a tool owner could configure it to setup their env and run their tests? Most of the tools are python, is there a python CI framework we should look at? Maybe easier to have a master script from the root dir that runs all CI for all tools? Just throwing some thoughts out.
Andrew
Quoted 4 lines
Cheers,
Andrew
Quoted 27 lines · Andrew Geissler
Quoted 6 lines · Andrew Jeffery
Quoted 5 lines
(5) How should I handle being in a subdirectory of a repository? The code I’m checking in will go into openbmc-tools. How does that work for CI?
Quoted 7 lines
There isn't any CI for openbmc-tools at the moment, as it was originally intended as a collection of quick hacks that made people's lives easier without any particular guarantees. This was done to lower the bar for entry and get people contributing their scripts.
CI is probably something we need to think about though, so I've added Andrew G in To/Cc:
I know for openbmc-build-scripts I did eventually make a jenkins job that does basic validation of the scripts it can (building docker containers). We were getting some regressions which were breaking our CI so this was a quick way to make sure the docker containers at least built when the scripts were updated.
We could do something similar with this repo, even if it's just running --help of the tools to ensure they execute. We could get more complex and connect them up to a QEMU session running openbmc to validate more. I don't have any bandwidth to do much here though so someone else would need to drive it.
Quoted 5 lines
Would we want a
run-ci.shin each subdirectory where a tool owner could configure it to setup their env and run their tests? Most of the tools are python, is there a python CI framework we should look at? Maybe easier to have a master script from the root dir that runs all CI for all tools? Just throwing some thoughts out.
The run-ci.sh approach might not be too bad - that way the jankins job can just run a find . -name run-ci.sh -exec \{\}\;, which feels kinda horrific but doesn't need much in the way of maintenance.
Andrew
Quoted 26 lines · Andrew Jeffery
Quoted 24 lines · Andrew Geissler
Quoted 11 lines · Andrew Jeffery
Quoted 4 lines
(5) How should I handle being in a subdirectory of a repository? The code I’m checking in will go into openbmc-tools. How does that work for CI?
There isn't any CI for openbmc-tools at the moment, as it was originally intended as a collection of quick hacks that made people's lives easier without any particular guarantees. This was done to lower the bar for entry and get people contributing their scripts.
CI is probably something we need to think about though, so I've added Andrew G in To/Cc:
I know for openbmc-build-scripts I did eventually make a jenkins job that does basic validation of the scripts it can (building docker containers). We were getting some regressions which were breaking our CI so this was a quick way to make sure the docker containers at least built when the scripts were updated.
We could do something similar with this repo, even if it's just running --help of the tools to ensure they execute. We could get more complex and connect them up to a QEMU session running openbmc to validate more. I don't have any bandwidth to do much here though so someone else would need to drive it.
Quoted 8 lines
Quoted 5 lines
Would we want a
run-ci.shin each subdirectory where a tool owner could configure it to setup their env and run their tests? Most of the tools are python, is there a python CI framework we should look at? Maybe easier to have a master script from the root dir that runs all CI for all tools? Just throwing some thoughts out.The
run-ci.shapproach might not be too bad - that way the jankins job can just run afind . -name run-ci.sh -exec \{\}\;, which feels kinda horrific but doesn't need much in the way of maintenance.
I would be interested in a run-ci.sh sort of thing as well for a repo I am working on. It would allow me to do some additional data validation with some custom scripts.
Quoted 1 line
Andrew