09:42 <hthiery> sjg1: hi, regarding the issue with buildman/binman (https://lists.denx.de/pipermail/u-boot/2021-October/464514.html). I would at least need a way to do a successful CI build. I think at this point it is required to have a dummy/fake binary blob for the signed_hdmi_imx8m.bin.

13:22 <sjg1> hthiery: Should we perhaps have a way to detect ext blobs and not run mkimage in that casee?

13:23 <sjg1> hthiery: Or I suppose we could get binman to create a dummy file in that case, so the mkimage works?

13:41 <hthiery> sjg1: at least a behavior like the one for the other missing blobs would be good ... I mean only throw a warning instead of an error

13:43 <sjg1> hthiery: OK possibly the easiest thing is to create fake blob-ext files in binman. That will work as long as mkimage isn't fussy as to the format, which is probably isn't since they are just binary blobs

13:46 <sjg1> Tartarus: I am trying to create a loop device in CI, with partitions: https://source.denx.de/u-boot/custodians/u-boot-dm/-/jobs/340213

13:46 <sjg1> Tartarus: but I get: mkfs.vfat: unable to open /dev/loop20p1: No such file or directory

13:46 <sjg1> I wonder if this is not supported, somehow?

13:47 <sjg1> Tartarus: the script is here: https://source.denx.de/u-boot/custodians/u-boot-dm/-/commit/94501c45410f2b08dc77e414a9fe1e4f72b34cf0

13:49 <hthiery> sjg1: creating a fake blob if it is not available would work. this is what I already tested (you can see that in my code snipped in the mail, there the fake was created by buildman)

13:50 <sjg1> hthiery: OK...do you feel up to implementing that in binman?

13:50 <hthiery> sjg1: I think adding an additional flag to the blob-ext could be checked for that

13:50 <hthiery> sjg1: I will give it a try and let you know .. ok?

13:51 <sjg1> hthiery: SGTM thanks

14:26 <Tartarus> sjg1: That's not something you can reliably do, I _think_ only containers with privileges can do that, and not all of ours run that way.

17:59 <sjg1> Tartarus: Should we change that, or is it a pain? I can probably check in a small compressed binary for now

18:00 <Tartarus> Can't change it, and should try and move to least privs required, but also see if that can be done without a huge huge time penalty

18:00 <Tartarus> ie when I tried moving to guestfish for FS stuff tests moved from seconds to tens of minutes, I think

18:01 <Tartarus> I almost wonder if we shouldn't just make a test objects repository

21:05 <sjg1> Tartarus: How come we can't change it? We seem to be able to mount images, just not detect the partitions, as if that feature is somehow missing

21:05 <sjg1> In this case the compressed test file is only 4KB so I can just include it, for now

21:07 <sjg1> Tartarus: hmm maybe the loopback partitions are somewhere else? I have heard of them being in a subdir

21:08 <sjg1> Tartarus: https://unix.stackexchange.com/questions/425639/loopdevice-partitions-do-not-show-up/425641

22:12 <Tartarus> You need to `losetup -P` for partition probing automatically, yes

23:05 <sjg1> Tartarus: Yes I have sudo losetup --show -f -P