Using the Yocto Project for Upstream Testing

Greetings All,

I've already mentioned this on the GCC side but it seems that the Yocto Project

has a class for testing upstream projects. The project is used for building embedded

distributions but due to the ability with bitbake and the amount of supported upstream

projects I've asked them to mention that other projects are interested in seeing this

happen at the usual Embedded Linux/IOT conference meeting they have yearly.

Just getting the LLVM side note that its already been discussed with them,

Nick

Sorry, I’m not sure I follow/understand your email, or what sort of replies/responses/discussion you’re hoping to get from it - perhaps you could rephrase and/or provide more detail?

David, Sorry its more to give you guys a heads up in case they ask about it being of interest and know that its something of interest outside that project. The Yocto and its related projects are a embedded build system. Since its goal is to build the whole custom embedded distribution I’ve suggested that they start testing upstream gcc/llvm. My reasoning is that while we have a testsuite for both having a software that builds and supports lots of real software can be of use. To my knowledge it has the ability to build over 10000 open source packages. Basically the idea is to automate testing of not just the testsuite but a suite of real software builds nightly or so as it takes time to build all that software. You can of course just build one piece of software and its dependencies separate through. This is a link to the project: And a link to all the recipes for building software it supports that are upstreamed: The GCC side knows as well but thought I would give you guys the heads up as well and sorry for not explaining it better, Nick

Ah, OK - thanks for clarifying. So the long and the short of it is that the LLVM project may receive bugs filed from this new source. Fair enough - so long as they’ve got good (ideally small, portable, etc) reproduction steps, hopefully they’ll be actionable/someone can prioritize the fix for them.

Yes that’s correct. Its more in the early stages of being implemented. There will be a talk at Embedded/IOT about planning for it and it may get mentioned as I’ve asked them to. If someone from this community does attend that conference I would suggest figuring out what you want in terms of working with them. Through I would note that reproduction steps may be hard depending on how many packages your building from outside using Yocto(this shouldn’t really be a problem due to Yocto being a set standard for distribution builds like make). If your really concerned about that you can find there list and ask about it there as well. Through most of the packages are basically upstream excluding some minor changes to the kernel, Nick