On 27 February 2013 11:38, Florian Fainelli <[email protected]> wrote: > On 02/27/2013 11:34 AM, Matthijs Kooijman wrote: >> >> Hi Florian, >> >>> I would rather we do not change this, as it breaks the following >>> workflow: >>> >>> - start the initial kernel build >>> - create a new patch using quilt in build_dir/target-*/linux-*/linux-*/ >>> - go through the round of building/testing/changing >>> - issuing make target/linux/refresh to copy back the patches >> >> It looks like these do not need to conflict. Rafał is talking about >> adding a patch to the package directory (e.g, /package/somewhere), while >> you're talking about adding a patch to the build_dir. > > > We are both talking about adding patches to the Linux kernel, not some > specific package, that case is already handled. If you update the timestamp > of a file in say, package/foo/patches/*.patch, your package foo will get > rebuilt, that does not work with the kernel, and I agree this can be > puzzling. > > >> >> For your workflow, it shouldn't hurt that a new patch in the package >> directory triggers a clean/prepare, right? > > > I may have to make any intermediate change to the sources, test them, re-run > quilt refresh for my patch to save them, but only when I am done, copy back > to target/linux/bar/patches-*/ either manually or using make > target/linux/refresh.
All this can be kept by doing it exactly like its done for packages: There you can disable this behavior through CONFIG_AUTOREBUILD), so it would make sense to add a CONFIG_AUTOREBUILD_KERNEL. Of course the autorebuild kernel will need a bit more intelligence as it needs to watch two patches* and files* directories, but it should be doable. Jonas _______________________________________________ openwrt-devel mailing list [email protected] https://lists.openwrt.org/mailman/listinfo/openwrt-devel
