On Tuesday 26 January 2016 13:35:04 Fengguang Wu wrote: > > There are a fixed set of config files for fast build tests (which I > > selected randomly, feel free to ask me to change the list to more > > reasonable ones): > > > > arm-allnoconfig > > arm-at91_dt_defconfig > > arm-at_hdmac > > arm-ep93xx > > arm-imx_v6_v7_defconfig > > arm-iop-adma > > arm-marzen_defconfig > > arm-prima2_defconfig > > arm-sa1100 > > arm-samsung > > arm-sh > > arm-spear13xx_defconfig > > And there is another set of best effort configs whose priority is > in-between the above list and the arch/*/configs/* ones. > > arm-arm5 > arm-arm67 > arm-mmp > arm-omap2plus_defconfig > arm-s3c2410_defconfig > arm-tegra_defconfig > I think we want at least one ARMv7-M NOMMU target in the list, I'd pick efm32_defconfig as it has the least overlap with the drivers in the other configs. I don't know what arm-sh is, but I assume it's a superset of marzen, so you can probably drop marzen or move it down to the low-prio set. Similarly, the at_hdmac is likely a subset of at91_dt_defconfig and could be dropped. prima2, ep93xx and spear13xx, mmp and s3c2410 are all slow-moving platforms, I would not expect to see much results from running those compared to the more active platforms. The platforms with the most changes these days are omap2plus, sunxi, shmobile, and exynos (samsung), followed by some that are widely used but not changed as much including tegra, imx_v6_v7, mvebu_v7, and qcom (previously msm). iop and sa1100 are rather old and won't change much, but it can be good to keep them in there because they are for older CPU architectures and behave a little different from the modern ones. It might be better to replace iop with ixp4xx, as this one is the only big-endian defconfig build we have, and that sometimes catches build-time bugs. Arnd