**** BEGIN LOGGING AT Wed Jul 19 03:00:03 2017 Jul 19 10:01:16 khem, wrt gcc6/7 I have rebuilt with -g and I am disassembling the kernels. Maybe I find something, I'll upload somewhere anyway Jul 19 10:02:53 there is a suspicious long long left shift somewhere thus __aeabi_llsl but is ignored by gcc7... Jul 19 10:03:38 no apparent ELF issues Jul 19 10:45:28 hi rburton , do you know if anyone is looking at the 1.12 gst patches? they've been on the list for some time. is there anything i can help with? Jul 19 11:04:35 khem, the shift is here: http://elixir.free-electrons.com/linux/latest/source/lib/xz/xz_dec_stream.c#L187 Jul 19 12:17:24 ndec: my big concern was the massive change to library depends which nobody replied to, then it turned out that video playback on x86 was totally bust too Jul 19 12:18:48 ouch. Jul 19 12:19:01 i will check the series. Jul 19 12:19:36 i had tried the first version of the patches that were on the list, and they were ok on my boards at least. Jul 19 16:35:05 Hmmm Jul 19 16:36:07 I'm going to start down the road of making a new task to do image compression, after image_complete, so that any of our images that are created can be compressed/converted Jul 19 16:36:37 image_fstypes already supports compression/conversion of any image type, what's the advantage to doing it after image_complete rather than before? Jul 19 16:38:18 kergoth: It doesn't support it for vmdk/qcow2/vdi Jul 19 16:38:24 nor I suspect the various live types Jul 19 16:38:55 And all of those types need a "regular" image made up first, usually ext4, but could be whatever the user selects Jul 19 16:39:11 as long as the filename is predictable, i don't see why we couldn't fix it to do so, rather than adding a second mechanism Jul 19 16:39:58 The problem I see first is we can't re-use CONVERSION_CMD_xxx Jul 19 16:40:09 Since they don't and can't use IMAGE_CMD Jul 19 16:51:58 kergoth: I assume a better idea didn't pop into your head yet, afk a few to find some lunch :) Jul 19 19:52:24 I might be able to whack this such that do_image_complete is still the end, but there's another task and so a few things like initramfs/etc that depend on do_image_complete wouldn't need tweaking Jul 19 19:52:30 * Tartarus keeps reading Jul 19 20:26:16 Hummm, crap, no, this won't work without a lot more re-jigging Jul 19 21:41:59 OK, back to plan 0, converting vmdk/etc to IMAGE_CMD was not as horrifying as I thought it might be Jul 19 21:43:46 $ git diff | diffstat -p1 Jul 19 21:43:46 meta/classes/image-vm.bbclass | 62 ++++++++++++++---------------------------- Jul 19 21:43:46 1 file changed, 22 insertions(+), 40 deletions(-) Jul 19 21:51:51 losetup -P on one of the files looks good Jul 19 21:51:56 will clean up and post soon I think Jul 19 21:52:03 need to test adding compression :) Jul 19 22:31:21 oh wait one problem left to solve, that's right, need to get the vmdk/etc dep on hdddirect working Jul 19 22:48:14 But, hey: Jul 19 22:48:19 $ ls tmp-glibc/deploy/images/qemux86/core-image-minimal-qemux86.vmdk* Jul 19 22:48:19 tmp-glibc/deploy/images/qemux86/core-image-minimal-qemux86.vmdk Jul 19 22:48:19 tmp-glibc/deploy/images/qemux86/core-image-minimal-qemux86.vmdk.bz2 **** ENDING LOGGING AT Thu Jul 20 03:00:01 2017