T42 | <FakeShell> mal: do you happen to have your libhybris vulkan icd code somewhere? i am trying to make an icd of vulkan and even tho now it picks up my icd, right after that vulkaninfo throws a Bus error. passing in VK_LOADER_DEBUG=all as an env var shows that it is finding my gpu and driver: | 02:48 |
---|---|---|
T42 | <FakeShell> Using "Mali-G68 MC4" with driver: "libvulkan_hybris.so.1" | 02:48 |
T42 | <FakeShell> to add to that, gdb shows no symbols or where we even are, as if the stack pointer is corrupted ^^ | 02:49 |
rinigus | @b100dian: nice to see you resuming :). as you have done already, droid-hal-pdx224-kernel and friends should go to droid-hal at obs | 06:43 |
rinigus | there should be nothing pdx223 specific in droid-hal-pdx223. we used the same approach for tama and it worked. | 06:44 |
rinigus | actually, we should probably make droid-hal-vendor-dlkm into pdx223 and pdx224 versions. | 06:46 |
rinigus | mal: probably dlkm can be somehow handled differently, but that way it allows to simply transfer los config to sfos | 06:48 |
rinigus | I am adding dlkm files to sfos root, not separate partition. that way its easy to update. | 06:48 |
rinigus | @b100dian: my mistake - dlkm is already different for pdx223 and 224 | 06:53 |
rinigus | @b100dian: for me, sb2 -t sony-xqct54-aarch64 rpm -qa | grep kernel does list kernel. kernel is needed to make sure that dlkm is the same version as the corresponding kernel. see spec buildrequires | 07:01 |
T42 | <b100dian> rinigus: thanks. I think I don't understand why just building droid-hal makes the dlkm package not be able to find the kernel one | 08:32 |
T42 | <b100dian> Maybe I need to do the other steps to up until --version | 08:33 |
T42 | <b100dian> But I remember an error at --config, let me try again | 08:33 |
T42 | <b100dian> hmm sb2-config showtarget: default target has not been set. Maybe its because of this | 08:40 |
rinigus | @b100dian: no, you should have the target set to as in sony-... well, if you copy commands from hadk | 08:58 |
T42 | <b100dian> I did set it with sb-config -d sony-xqcq54-aarch64 and rebuilt packages with —droid-hal .. | 09:12 |
rinigus | so, for some reason the package is not pulled via buildrequires from local repo. no idea why. also, not sure it was pulled automatically in my case either | 09:25 |
rinigus | @b100dian: I can reproduce your error, I think. it looks to be an issue for the first build, when you don't have yet kernel installed | 09:55 |
rinigus | now somehow it was installed for me before (I removed it from repo as a part of the test) | 09:56 |
T42 | <b100dian> So, to proceed with the next steps (config and version) I would like to know if you have local changes to build --config with HABUILD_DEVICE vs DEVICE diffs? | 10:13 |
rinigus | @b100dian: I don't build them locally, only via OBS | 10:33 |
rinigus | how did you build dlkm? | 10:33 |
T42 | <b100dian> `rpm/dhd/helpers/build_packages.sh --build=hybris/mw/droid-hal-vendor-dlkm --spec=rpm/$HABUILD_DEVICE.spec` is the command that fails | 11:05 |
T42 | <b100dian> `rpm/dhd/helpers/build_packages.sh --droid-hal` before that succeeds | 11:05 |
T42 | <b100dian> it might be that you have build the whole thing and _then_ added the dlkm dependency in the dhd | 11:06 |
rinigus | @b100dian: if you set mainver and relver in spec to some numbers and after the first build swap to the current implementation it would work too. I'll try to find a better working solution | 11:08 |
T42 | <b100dian> Good, let me try that next - right now I figured the manifest didnt sync hybris/droid-configs :D so I am trying to build forward up to —version | 11:09 |
rinigus | locally? | 11:11 |
T42 | <b100dian> too much error prone ;) ? | 11:12 |
rinigus | @b100dian: no need as we have obs anyway. | 11:18 |
rinigus | and keeping local config minimal | 11:18 |
T42 | <b100dian> your last suggestion worked, I've uploaded the dlkm package, thanks | 11:20 |
T42 | <b100dian> I am just going out, will test in the evening. | 11:20 |
rinigus | enjoy! | 11:21 |
rinigus | hmm, in principle we have already kernel-modules. which probably can be loaded in linux manner, not using scripts in android side. something that maybe even nicer in longer term | 11:28 |
rinigus | @b100dian: found the way to install kernel before building dlkm. updated our hadk accordingly in dlkm section | 11:47 |
rinigus | @b100dian: automatically updated vendor_boot with the wrong dtbo and phone refused to boot. on top of that, was unable to access the phone anymore via fastboot and even newflasher got "Error, no DATA reply!" | 13:22 |
rinigus | fortunately, emma managed to flash stock android 13. so, lots of updates coming up, I guess | 13:23 |
rinigus | its tempting to remove vendor_boot and boot img flashing from configs via update scripts. and let flash the images manually. but hopefully its isolated case. | 13:24 |
T42 | <b100dian> Hmm I think I got that no DATA and used flash mode for newflasher... | 13:24 |
rinigus | @b100dian: I was in flash mode with newflasher (green led) | 13:25 |
rinigus | spooky | 13:25 |
T42 | <b100dian> Yeah, but I definitely didnt have a windows machine for emma, so it was either green or fasboot blue led | 13:26 |
T42 | <b100dian> So the uodates will be tested on devel by us :) | 13:26 |
rinigus | before newflasher worked, its the first time for this | 13:30 |
rinigus | @b100dian: was it that after unlocking bootloader, system updates should be done via newflasher? | 13:41 |
T42 | <b100dian> rinigus: if you mean that the stock OS won't notify/install sony updates I guess that's correct. Not sure about emma. But mono xperifirm + newflasher should work, though it's a hassle to back up just to reflash everything | 14:22 |
T42 | <b100dian> maybe extracting xperifirm sin files with unsin and flashing them discretionary | 14:22 |
T42 | <b100dian> or even better, re-running extract-proprietary blobs and re-building lineage vendor..? | 14:23 |
T42 | <b100dian> Fantastic job, rinigus! It booted right away : https://irc.thaodan.de/.imgstore/4b609898/file_7501.jpg | 15:08 |
rinigus | @b100dian: nice! happy it worked. I'll later will proceed with the update of stock via newflasher (files are already here), then to LOS, and then to SFOS. as all files are in place, shouldn't take too long. if it works... | 15:23 |
rinigus | as for emma, that did not provide the latest stock version for some reason. | 15:23 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!