10:27 <paulbarker> Hi folks, I'm trying to reproduce the failure in https://source.denx.de/u-boot/u-boot/-/jobs/316664. buildman is giving me the error that I have no tool chains

10:28 <paulbarker> The suggested command `buildman --fetch-arch all` installs far too much, I killed it after it had pulled several GB of toolchains

10:28 <paulbarker> Is there a way to find out which toolchain is needed to build sandbox_spl so I don't need to install all of them?

10:36 <paulbarker> I can build following the instructions in `doc/arch/sandbox.rst` but the command in the test log `./tools/buildman/buildman -T0 -o ${UBOOT_TRAVIS_BUILD_DIR} -w --board sandbox_spl` fails

10:45 <LeSpocky> is there a way to ask for confirmation in the u-boot console when running an u-boot script with potential destructive actions?

10:56 <LeSpocky> there's confirm_yesno() in include/console.h but it's not exposed to any command for scripting

11:10 <milkylainen> LeSpocky: No hush shell read?

11:12 <milkylainen> https://github.com/u-boot/u-boot/blob/master/common/cli_hush.c#L517

11:55 <LeSpocky> well I did not look for builtin commands ^^

11:55 <LeSpocky> thx

12:06 <milkylainen> LeSpocky: np

12:23 <sjg1> paulbarker: For sandbox you only need the native toolchain for your machine: sudo apt install build-essential

12:23 <sjg1> paulbarker: But see the documentation for building U-Boot which has all of this

12:24 <sjg1> paulbarker: For this one: binman.ftest.TestFunctional.testFullHelpInternal testtools.testresult.real._StringException: Traceback (most recent call last):

12:24 <sjg1> paulbarker: you need to run ./tools/binman/binman test

12:25 <sjg1> paulbarker: so 'buildman sandbox -o somedir

12:26 <sjg1> Then ./tools/binman/binman --toolpath somedir/tools test

12:30 <Tartarus> paulbarker: It's also possible you'll just want to grab the container and re-run the tests in there instead, if you're still having a problem. OTOH I see you posted v3 and v4, so maybe you got it tested :) I'll try and fire off something shortly since I recall it was kind of annoying to reproduce the failure I saw, but it was repeatable in CI at least.

12:46 <paulbarker> Tartarus: I figured out I could add a toolchain alias of 'sandbox' -> 'x86_64', install that single toolchain and build

12:47 <paulbarker> Couldn't run all the tests as `lzma_alone` isn't packaged for opensuse as far as I can find, but I could re-run the test that failed, see the failure then fix it

12:52 <Tartarus> OK, good :)

13:19 <Tartarus> paulbarker: And reading the changelog now, yeah, wow, that's some python craziness

13:41 <milkylainen> sjg1: nice! I'll take a look as soon as I get some spare time. Any defconfig changes I need to enable to get support for the uefi protocols? Or are they just there in app mode?

16:15 <mrnuke> Tartarus: I saw the PPC build fails on gitlab. I'm looking too see what fails

16:25 <Tartarus> mrnuke: I've got that, it's FSL_CAAM needs to imply SPL_CRYPTO if (ARM && SPL)

16:26 <Tartarus> https://www.irccloud.com/pastebin/O2jkgA5R/

16:26 <Tartarus> That's more of a wtf I'm trying to figure out right now

16:28 <Tartarus> Ah, I think I see, problem in common/hash.c

16:28 <Tartarus> #ifdef USE_HOSTCC

16:28 <Tartarus> #endif

16:28 <Tartarus> #else

16:28 <Tartarus> # define I_WANT_MD5 IS_ENABLED(MD5)

16:28 <Tartarus> # define I_WANT_MD5 1

16:29 <Tartarus> But IS_ENABLED() needs CONFIG_foo, CONFIG_IS_ENABLED(foo) is what should be there

16:29 <Tartarus> I can fix that and re-run

16:40 <mrnuke> Tartarus: I'm sorry this series causes you so much headache. I wish it had just worked (forgot to builbman powerpc).

16:40 <Tartarus> no worries

16:41 <mrnuke> Tartarus: I do appreciate your help with this series. You did save me a lot of time. Thank you!

20:03 <Tartarus> mrnuke: OK, so here's a concern. It seems like that FIT_SHA256 select'd SHA256 meant that sha256 was a valid option for FIT verification before, but is dropped by your series

20:05 <Tartarus> Or maybe the users weren't really using it anyhow, looking at others.

20:19 <mrnuke> Tartarus: I did consider that. I would expect defconfigs to select FIT_SHA256, but they don't. There wasn't any other config that "select FIT_SHA256"

20:24 <mrnuke> I see your point with "default y" -> FIT_SHA256 -> "select SHA256"

20:26 <mrnuke> I suppose it wouldn't hurt for "config FIT" to "imply SHA256" I can test this locally to see if it builds

20:29 <mrnuke> Tartarus: you're already ahead of me on that too! Reminds me of that commercial where some detectives were tracking and AMD GPU, and couldn't keep up with it because "It's too fast!"