/srv/irclogs.linaro.org/2018/06/27/#linaro-tcwg.txt

ramanamaximk, good morning ..09:08
maximkramana: good morning!10:02
yrouxmaximk, MarkMurr: Do we do a release sync call today ?10:12
maximkyroux: I don't think we have much to discuss.10:13
maximkMarkMurr: ??10:13
yrouxmaximk: that's my feelings as well10:14
MarkMurryroux: I don't think so. Next release cycle will be around August/Septermber/October. I'll do 7.4 and 6.5 when those release are done and the branch merges are finished. Until then there is not much to talk about :-)10:14
yrouxMarkMurr, maximk: ok for me10:15
MarkMurryroux,maximk: 6.5 ought to be the final 6.* release, and after the above, there will be 2 more 7.* releases, both next year.10:16
maximkyroux: MarkMurr: ok to leave just Wednesday call and make it bi-weekly?10:16
yrouxmaximk: ok10:17
MarkMurrmaximk,yroux: OK - and even then, only have them when there is something to talk about?10:18
maximkMarkMurr: ok.10:18
MarkMurrmaximk,yroux: So next try in 2 weeks time? Maxim - please reschedule on the calendar and re-invite me.10:19
maximkMarkMurr: yroux: invite updated10:20
yrouxMarkMurr, maximk: I'll be on vacation in 2 weeks, but feel free to have a call without me ;)10:21
MarkMurryroux: I think I won't make the next two calls as I'll be busy with conferences and stuff.10:28
MarkMurrI just got a HUGE rush of email backlog into my linaro mailbox; looks like about a weeks' worth, or maybe a bit more.10:30
yrouxMarkMurr: greylisting !?10:33
MarkMurryroux: Dunno. Not by me.10:33
yrouxMarkMurr: ok weird... in one of these email I think you should comment on LDTS about missing release symlinks10:34
MarkMurryroux: from Sumit Garg? I can't see what is wrong?10:41
MarkMurrOr has it already been fixed?\10:42
MarkMurr..oO(looks like it)10:42
yrouxMarkMurr: it was more to answer to concern danielt and take care that when the deployment ticket is done it is written that symlinks must be updated accordingly10:48
MarkMurryroux: Aaah, OK. I'll add that to my notes.10:49
danieltMarkMurr, yroux: I believe the ticket got escalated outside of the support channels...10:51
danieltMarkMurr, yroux: Last I saw it looked like they were concluding that this manual process is too error prone (latest/ has been mismanaged in 3 of the last 3 releases... that's probably a sign of a process problem).10:52
MarkMurrdanielt: I was unaware of the symlink concerned, so some of the faut lies with me in raising the IT ticket to do the job.10:53
yrouxdanielt: ok10:56
danieltMarkMurr: I'm about to reply to Bill on-ticket... however it will likely stay open until some process change is enacted (whether that is your notes or fully automating the latest links).10:58
MarkMurrdanielt: OK. I have three more releases to go on the 7.* branch. It shouldn't be too hard to get this right at IT-ticket time.11:01
ramanaIs there a call today ?12:01
ramanamaximk, MarkMurr ^^12:01
maximkramana: no12:01
ramanamaximk, something popped in my calendar - so I thoguht I'd check12:02
maximkramana: and I've sent an updated invite to have call every other week12:02
maximkramana: only on Wednesdays12:02
* ajb-linaro tries armie13:21
* ajb-linaro is dissapointed 13:31
ajb-linarosadly armie fails in it's interactions with the kernel13:31
ajb-linarohence RISU won't work13:31
druei'm looking to build linux 4.17 and mainline using clang, but i understand it may be broken due to asm goto.. is there any workaround or update on this?15:36
arnddrue: the asm-goto stuff is only needed on x86 I think15:37
arndwould it be enough to start with arm64?15:37
drueoh we're only targeting arm64 actually.15:38
arnddrue: ok, I just checked that doesn't need asm-goto yet15:38
druethat's great news. the other question is on choosing which toolchain to use. we're planning on using aosp's, for lack of something more upstream15:39
drue(background: kv-93 in jira)15:39
arnddrue: I think you need a fairly recent compiler to build the kernel15:41
arndif the aosp version of clang is anything like their gcc, it's completely outdated and buggy15:41
arndI'd start with the 6.0 release15:41
drueand patched15:41
druebut it works.. but i'm concerned about false positives for issues they're patching15:42
druearnd, i'd like to not build the toolchain, does anyone distribute it?15:42
arnddrue: I don't know if we provide any binaries on linaro.org, have you looked?15:44
arndOn https://releases.linaro.org/components/toolchain/binaries/ I only see gcc based toolchains, but maybe I'm looking in the wrong place15:45
psmithdrue: There are binaries available on http://releases.llvm.org/ If you are going to be cross-compiling from x86 you'll need the x86 binaries. The aarch64 ones are aarch64 native.15:45
druei was told we don't.15:45
arndpsmith: are you sure those binaries are not multiplatform-targetted?15:46
arndwhen I used clang in the past to build kernels, I could always use the same binary for x86, arm32 and arm6415:46
drueok i'll give those a shot starting with latest (6.0.1)15:47
psmithBy aarch64 native I mean the executable will only run on aarch64.15:47
psmithBeware that the clang distributions are not complete toolchains. I suspect that you'll need to point clang at a gcc toolchain to pick up libc libgcc etc.15:48
psmithFor the linaro toolchains installed in /path/to/linaro-toolchain then I recommend using --gcc-toolchain=/path/to/linaro-toolchain and --sysroot=/path/to/linaro-toolchain/aarch64-linux-gnu/libc15:50
psmithTo target aarch64 you'll need at least --target=aarch64-linux-gnu15:50
arndpsmith: the kernel passes the correct --target argument to clang when cross-compiling15:51
psmithapologies I've not tried to compile the kernel.15:52
psmithOne potential issue I've seen fly with clang is that there are some gcc command line options used that clang doesn't support when CONFIG_ARM64_LSE_ATOMICS=y is used (https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/lib/Makefile#n15) the options are -ffixed-(register) and -fcall-saved-(register)15:58
psmithI don't know whether that will be an issue for you, but worth looking out for.15:58
arndpsmith: we can now make kernel config options depend on toolchain capabilities, it sounds like someone should send a patch to make ARM64_LSE_ATOMIC depend on !clang. let's see what other issues we run into first though ;-)16:01
arnddrue: I've managed to build simple files using the clang binaries from llvm.org and an appropriate --gcc-toolchain location, but still fail  to build the kernel on my machine. I'll try a little more and let you know if I find a solution. It's probably not hard16:02
arnddrue, psmith: I got it working now, the only thing I needed was to add two symlinks in the gcc cross toolchain16:10
arndmy gcc was configured for "aarch64-linux", while clang was looking for "aarch64-none-linux-gnu"16:11

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!