**** BEGIN LOGGING AT Wed Oct 16 02:59:57 2019 Oct 16 07:32:23 karlp: I was on rpi3 or rpi4 and selected the appropriate firmware but it required another firmware package that was not at dependency of the rpi3 or rpi4 firmware Oct 16 07:32:49 let me find it Oct 16 07:33:44 karlp: see pm Oct 16 07:44:19 mornin' Oct 16 07:44:59 ldir: o/ Oct 16 07:45:06 how did the dogfooding end :) Oct 16 07:46:26 he he - ok in the end - but I have a shell escaping question. Oct 16 07:47:15 shoot Oct 16 07:48:34 I'm fiddling with an existing script, trying to tidy it up. It does some horrible things - If anyone remembers 'beardropper' then you'll understand where I'm starting Oct 16 07:49:56 yeah, it looks a bit eye-cancerous Oct 16 07:50:30 it has some truly horrible syntax e.g. some stuff... | tr , \ ) When you're escaping spaces it's truly horrible Oct 16 07:51:13 so I think that tr ',' ' ' is an improvement Oct 16 07:51:29 you can often avoid space escaping entirely be changing $IFS as needed, e.g. to newline only Oct 16 07:52:35 also depending on what you need to substitute exactly, the extra fork/exec for tr can be avoided as well Oct 16 07:53:06 e.g. foo="$(echo "$foo" | tr ',' ' ')" => foo="${foo//,/ }" Oct 16 07:53:29 ${var/xxx/yyy} -> replace xxx with yyy once in $var Oct 16 07:53:41 ${var//xxx/yyy} -> replace all xxx with yyy in $var Oct 16 07:53:55 I'll look at that in round 28 of the tidy up :-) Oct 16 07:55:07 so the escaping question - I have a tr that's turning ' into space. it was coded as tr ' \ (escaping the space) Oct 16 07:56:29 I now have it as tr \' ' ' - but I'm trying to keep to a consistent standard of having the character/s quoted. Which i think means tr '\'' ' ' Oct 16 07:56:39 but the \ escape doesn't appear to work Oct 16 07:56:56 yes. '...' is special in shell Oct 16 07:57:04 am I being mental/OCD? Oct 16 07:57:11 it does not allow escaping ' for security resons Oct 16 07:57:39 the only way to escape ' with '...' is to close the string first and reopening it afterwards Oct 16 07:58:02 so foo'bar needs to be written as 'foo'\''bar' Oct 16 07:58:32 or in your case if it is literal code, you might prefer "foo'bar" or "'" for your specific tr example Oct 16 07:58:50 ha - so tr ''\''' ' ' :-) Oct 16 07:58:58 exactly Oct 16 07:59:03 or tr "'" ' ' Oct 16 07:59:17 which is more obvious in almost all cases imho Oct 16 08:00:52 so when they say ' ' quotes don't do expansion they mean it doesn't do anything at all - it's a literal Oct 16 08:01:07 right Oct 16 08:01:35 which is a powerful facility when you want to prepare safe code for eval Oct 16 08:02:04 excellent - thank you jow - expert as always! Oct 16 08:02:08 all you need to do is to ensure that all ' in the input are escaped as '\'' then enclose the result in '...' and its ready for eval without unintended interpolations Oct 16 08:02:39 which becomes useful when you want to emulate array like operations without native array support Oct 16 08:03:09 anyhow, I'm getting offtopic Oct 16 08:05:06 ah yes eval - I'm 'learning' about that - yesterday I managed to work out that by setting a variable to "foo;ps" I was able to get an eval to unexpectedly run a 'ps' Oct 16 08:06:05 btw if the input and output of your tr is a variable (like var=$(echo $var | tr ...) youmight want to use variable substitution instead Oct 16 08:06:21 beware unexpected ';' Oct 16 08:06:39 var="${var//\'/ }" will be equivalent to var="$(echo "$var" | tr ''\''' ' ')" Oct 16 08:07:21 without the overhead of a subshell and executing tr Oct 16 08:08:12 I'll look to see where I can use that trick Oct 16 08:08:42 I wonder why ' is substituted in the first place Oct 16 08:11:57 The bit of code I'm looking at parses the output of 'set' which single quotes the values of variables and we don't want that Oct 16 08:15:41 * ldir goes for breakfast - not dogfood Oct 16 08:42:55 ah I see, well stripping single quotes is not actually enough I think Oct 16 08:43:11 because the 'set' output could contain escape sequences etc. as well Oct 16 08:43:22 mornin' all o/ Oct 16 08:43:25 so the right hand side of set should be eval'd to get the proper value Oct 16 09:44:07 pkgadd: it does for brcmfmac, but not for the bluetooth bits and the clocks for them on sunxi, aiui. Oct 16 09:50:08 well that's a PR sent into the dropbear repo Oct 16 10:13:26 https://git.openwrt.org/?p=openwrt/staging/ldir.git;a=commit;h=8b230e2e7e178c80e046c53e0de896a692b19a8b Makes sense right? Oct 16 10:18:17 at ldir at least I cannot claim that it does not make sense Oct 16 10:18:23 *ldir: ^ Oct 16 10:28:43 lol Oct 16 10:29:29 s/KMAKE/HOST_KMAKE/ (my wednesday bikeshed) Oct 16 10:33:44 yes I like that - even more obvious Oct 16 10:44:22 ynezz: https://git.openwrt.org/?p=openwrt/staging/ldir.git;a=commit;h=932361880ffeddc991740d35282c01b9b43dd7c5 better? Oct 16 10:44:40 and it still compiles which is a result :-) Oct 16 10:46:07 ldir: to me it doesn't really make sense Oct 16 10:46:29 the even for the target kernel, the kernel build system makes a distinction between host cflags and target cflags Oct 16 10:46:46 for the kernel headers it will build some host utilities, yes Oct 16 10:47:09 but isn't CFLAGS supposed to refer to the target cflags only anyway? Oct 16 10:47:31 and the headers it emits are meant to be used in code that is built for the target, not the host Oct 16 10:47:53 it probably doesn't make a huge difference, since it shouldn't really build any target target utilities at this point anyway Oct 16 10:48:54 looking at the target kernel make command, it doesn't seem to set CFLAGS at all Oct 16 10:49:42 so maybe what would make more sense (and be in line with your commit description) is if you replace CFLAGS with HOSTCFLAGS in that line (to match up with the corresponding line in KERNEL_MAKE_FLAGS from kernel.mk) Oct 16 10:49:46 does that make sense? Oct 16 10:50:03 I've seen the host utilities built at this stage be built with the target cflags - which didn't make sense in that some of the flags were only applicable to the target. (eg -march-btmach Oct 16 10:50:47 odd Oct 16 10:51:50 and it prompted me to look at why scripts/kconfig/conf was being built with TARGET_CFLAGS - and that's were I found the smoking gun Oct 16 10:52:22 then again, using CFLAGS for the override might be the wrong variable in general. can you test if setting HOSTCFLAGS makes the host cflags appear for the host utilities in that step? Oct 16 10:54:27 sure, will look later on - remember I'm no expert, I just stumbled across this and said 'hmm, that's wrong'. There may well be a more correct fix Oct 16 10:56:45 and I stumbled across it as part of the building on Catalina fallout Oct 16 14:22:59 nbd: I think your HOSTCFLAGS is the correct thing to do https://git.openwrt.org/?p=openwrt/staging/ldir.git;a=commit;h=7d2afe93c0c38eea591f3a461b19d32802b75019 - https://pastebin.com/tnQFVbCV Oct 16 14:31:06 * ldir adds a bit more green Oct 16 14:44:40 ldir: looks good to me Oct 16 14:47:23 nbd: thank you - have done something right today \o/ Oct 16 15:16:36 * ldir has managed to remember how to push a commit Oct 16 15:18:35 * ldir waits for the buildbots to fall over ;-) Oct 16 15:34:11 nbd: ping Oct 16 15:38:54 blogic: still nothing about the 802.11r/apple issue? Oct 16 15:39:03 DonkeyHotei: nothing, sorry Oct 16 15:39:09 i did nudge them today Oct 16 15:39:14 thanks for trying Oct 16 15:39:56 as an apple owner/user - what is the issue and who is them ? Oct 16 15:40:19 ldir: regression in mac80211 breaking 11r that was apparently fixed Oct 16 15:40:50 we know there is a problem in our latest and i played chinese post with ppl and was told that there is a fix and asked for details Oct 16 15:40:59 so the fast roaming is less than fast ? Oct 16 15:41:39 or not at all Oct 16 15:42:06 i dont use hipster tech so it never worked for me from the start :-) Oct 16 15:42:30 * ldir holds it wrong and doesn't care :-) Oct 16 15:45:34 i've actually given up on using 5GHz with my Apple devices as the latency spikes & packet loss are silly silly high. Tried mt76 as well and that exhibited same problem. Of course common aspects... Apple & Linux APs. Need to fire up a non Apple 5GHz device and test that. Hmmm. Oct 16 15:46:21 and/or boot macbook into Windows natively and see if issue persists. Hmmmm. Oct 16 15:56:13 or move to smaller flat/house so you don't need 802.11r :p Oct 16 16:38:11 how curious - boot the same hardware into Windows and wouldn't you know... latency & packet loss issues go away. Oct 16 16:40:05 ynezz: Even with one AP, if it is a dual-band AP, 802.11r can be useful. That said, so many devices have broken FT-SAE support that I had to disable 802.11r anyway. :( Oct 16 16:40:25 (FT-SAE is 802.11r for WPA3) Oct 16 17:30:31 blogic: pong Oct 16 17:35:53 oooohh lookit https://github.com/mkj/dropbear/pull/83 "I like this" - he's wise to try it first though :-) Oct 16 18:01:03 hello, does anyone offer commercial support for OpenWRT? Oct 16 18:01:30 I've got an issue with OpenWRT and Quectel LTE modems that I'd like to hire someone to fix for me Oct 16 19:20:02 finally bit the bullet and ordered myself a new ryzen system Oct 16 19:20:25 with 64g ram to build openwrt entirely in tmpfs Oct 16 19:20:42 all these toolchain rebuilds are getting annoying Oct 16 19:21:20 =) Oct 17 00:17:41 ryzen systems make compiling openwrt much faster with all that l3 cache **** BEGIN LOGGING AT Thu Oct 17 02:28:05 2019 **** ENDING LOGGING AT Thu Oct 17 02:59:57 2019