**** BEGIN LOGGING AT Sun May 23 02:59:56 2021 May 23 16:54:13 updated openwrt/upstream, https://sdwalker.github.io/uscan/index.html May 23 23:07:18 So turns out, I have the right version -- version.buildinfo appears to use the commit before, c.f. https://downloads.openwrt.org/releases/21.02.0-rc1/targets/ath79/generic/version.buildinfo May 23 23:09:11 However, my kernel version is wrong; I get 5.4.111-1-add660cce6f0585da900a76dc3e37873 while OpenWrt wants 5.4.111-1-4f1271a56f0192cef6cc5531235d2d0e (https://downloads.openwrt.org/releases/21.02.0-rc1/targets/ath79/generic/packages/Packages) May 23 23:09:29 so I'm guessing I need to be more careful about updating package feeds May 23 23:17:59 hurricos: the version number includes the hashed kernel configuration, so any change at all in that regard will change the ABI dependency May 23 23:19:14 slh: So that chunk /111-1-(.*)$/ at $1 is the vermagic? May 23 23:19:29 sorry for the perlism, I just mean, 4f127* May 23 23:19:34 if you want to reproduce the release builds, you probably need to build all kernel modules (so that the right features are enabled in the kernel) May 23 23:19:46 CONFIG_ALL_KMODS in https://downloads.openwrt.org/releases/21.02.0-rc1/targets/ath79/generic/config.buildinfo May 23 23:20:11 add660cce6f0585da900a76dc3e37873 is, the stuff before is just the upstream kernel and package release May 23 23:20:23 Very nice. May 23 23:21:29 OK, I'd recalled 'vermagic' from documentation, that helps. Adding CONFIG_ALL_KMODS, is there any other obvious kernel-related top level there May 23 23:21:30 ? May 23 23:22:26 I'd guess CONFIG_ALL_KMODS is enough actually just because which modules you build concurrently oughtn't directly change the vermagic May 23 23:22:33 just the features May 23 23:22:41 that *those* seleect May 23 23:22:49 anyways, worth a shot May 23 23:22:56 you probably also have to build for all devices (of your target), as those might enable kernel options May 23 23:25:19 unless you go full buildbot config (enabling lots of 'crap' you wouldn't need yourself), you will have to experiment a bit to get a matching ABI hash for the kernel dependencies (how much is enough, before going all in and waiting hours for a build and using up ~100 GB of build space) May 23 23:25:24 It really does sound like we should be hosting our own package remote. May 23 23:25:39 Rather than reusing OpenWrt's May 23 23:26:27 I'm not entirely sure why this worked previously; I hesistate to say we might actually have never been installing any kernel modules to notice this. May 23 23:26:31 I would be surprised that enabling devices would change the kernel config if you already have CONFIG_ALL_KMODS May 23 23:26:47 well let's take a shot with CONFIG_ALL_KMODS. See y'all in 26 minutes :^) May 23 23:26:50 and thank you! May 24 00:14:35 zorun: CONFIG_ALL_KMODS gets me 2d82b9d89a8772b937ba9bc1f527eeb2. :^) May 24 00:17:12 I think I've found the other config calling the issues though May 24 01:08:42 one more attempt gives me efb4aeea915eb905cb7e58820c3294a5. I'm just going to use https://downloads.openwrt.org/releases/21.02.0-rc1/targets/ath79/generic/config.buildinfo directly :| May 24 01:13:09 Unless someone wants to look at this diffconfig and suggest what nonstandard kernel configs this'd trigger? https://paste.c-net.org/CacheHelen May 24 01:17:37 I didn't even realize OPENSSL stuff was being re-selected, I bet that's a part of it too. **** ENDING LOGGING AT Mon May 24 03:00:16 2021