On 2018-08-19 12:30 PM, Mathias Kresin wrote: > 08/19/2018 05:46 PM, Chuanhong Guo: >> Another difference there is that eth0 and eth1 are swapped for chips >> with builtin switch (except ar7240). And I think this one makes it >> really annoying to write a config migration script. > > Indeed, it will be a pain. Not that I really like the idea, but > something like > https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/layerscape/base-files/lib/preinit/05_layerscape_reorder_eth > comes in mind as a "fix". >
FYI I was NAK'ed by John Crispin (blogic) on such a thing in: https://github.com/openwrt/openwrt/pull/1230#issuecomment-409213217 (his reponse was https://github.com/openwrt/openwrt/pull/1230#issuecomment-409218211 >> We now have some boards got SUPPORTED_DEVICES from ar71xx and some >> boards not. I'm confused about whether we should add such a >> SUPPORTED_DEVICES string when we port a board from ar71xx to ath79. >> (And I agree that my patch here didn't improve this situation.) >> I hope we could either add all the missing ones or remove all the >> existing ones. > I agree. I think we should have SUPPORTED_DEVICES but that is a mild preference vs. strong feeling. > I like to see an agreement on this topic as well. For now I accept > what the contributor considers the way to go is. > >> I wanted to make this RFC on mailing list but then I >> think this discussion will end up nowhere :( >> >> So...This patch can be dropped as it improved nothing... > > Marked as rejected > > Mathias > >> Dmitry Tunin <hanipouspi...@gmail.com> 于2018年8月19日周日 >> 下午11:40写道: >>> >>> вс, 19 авг. 2018 г. в 17:46, Mathias Kresin <d...@kresin.me>: >>>> >>>> 2018-08-19 15:47 GMT+02:00 Chuanhong Guo <gch981...@gmail.com>: >>>>> These lines are coming from ar71xx to allow using sysupgrade to >>>>> switch from ar71xx to ath79. But a sysupgrade with config preserved >>>>> won't work since some of the config files are incompatible. >>>> >>>> To be honest, I don't see that your patch really fixes the issue. Even >>>> if you drop the ar71xx compatible string, it's possible that people >>>> are using a forced sysupgade and therefore have the same problem >>>> again. Means, it's rather a "might work" workaround. Furthermore, >>>> there aren't only tp-link boards affected by this issues. I would >>>> really like to see a treewide handling of the issue. >>>> >>>> It isn't that uncommon that something changes and an upgrade of >>>> existing user configs is required. We're usually add uci-defaults >>>> scripts to do so. One example of doing so can be found in the lantiq >>>> target[0]. >>>> >>>> I'm not yet in a position to say what the correct approach would be >>>> here. I'm only aware that the "option path" in /e/c/wireless has >>>> changed for some(?) boards. No idea what else has changed between >>>> ar71xx and ath79. >>>> >>> Frankly speaking even this path change doesn't hurt. If you upgrade >>> from ar71xx to ath79 with a wrong (for ath79) path, >>> new entries for wireless devices are added to /etc/config/wireless >>> with correct path. >>> >>> I upgraded a lot these days on different devices from ar71xx to ath79 >>> and back with keeping configs. >>> Nothing really wrong happens except a few useless lines in >>> /etc/config/wireless. >>> >>> Even if this happens the correct wifi device will be disabled because >>> of the default config. >>> In this case user will open the file and see what happened. >>> >>> So I don't see any sufficient reason to prevent upgrading with the >>> old configs. > > > _______________________________________________ > openwrt-devel mailing list > openwrt-devel@lists.openwrt.org > https://lists.openwrt.org/mailman/listinfo/openwrt-devel _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel