github: workflows: test: build only on windows, test on cm3 - #91
Conversation
that is too general, only the cortex_m0 is a problem right now, so you can also just skip that and allow the others to continue building/running. Whatever makes sense.... |
zephyrproject-rtos/zephyr#112124 |
|
Another option could be to limit the number of concurrent Twister jobs involving these QEMU targets on MS Windows to 1 (suggested by @nashif). This works for zephyrproject-rtos/example-application#105 |
Feels like a hack to be honest but we need to come up with a coherent story between the repository, right now it's just a big mess. Maybe we can run with concurrency one on main (which tests on every push) and build only on the other repo? |
|
My main fear with the concurrency trick is that we end up in a situation where it works sporadically depending on how loaded the github runner is. |
That's a fair concern. |
| PLATFORMS_FLAGS+="-p native_sim " | ||
| fi | ||
| PLATFORMS_FLAGS+="-p qemu_cortex_m0 " | ||
| PLATFORMS_FLAGS+="-p qemu_cortex_m3 " |
There was a problem hiding this comment.
Not sure why the m3 platform was added? Did you mean to remove the m0?
There was a problem hiding this comment.
My understanding is that it's more reliable, undecided on whether to drop m0 or not.
There was a problem hiding this comment.
Let's drop it, whatever.
As of recent discussion it seems like qemu on windows is unreliable run those tests as build only and test on qemu_cortex_m3 which should be more reliable overall. Link: zephyrproject-rtos/zephyr#113543 (review) Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
|
Not that it's particularly reliable even with build only https://github.com/fabiobaltieri/action-zephyr-setup/actions/runs/29823824223/attempts/1 |
3 participants
As of recent discussion it seems like qemu on windows is unreliable run those tests as build only and test on qemu_cortex_m3 which should be more reliable overall.
Link: zephyrproject-rtos/zephyr#113543 (review)