**** BEGIN LOGGING AT Thu Sep 28 03:00:02 2017 Sep 28 05:08:44 Tartarus: I have seen wrong call stacks with gdb on mips, but not on other arches Sep 28 05:10:44 is it ppc ? Sep 28 07:02:37 Crofton|work, bmouring, thank you! I see the reasons. Anyway, I saw the potential gap and a way out: I created two recipes, that virtually provides the same functionality. I just have no way to select one or other in image recipe. That's pity. Sep 28 07:17:19 good morning Sep 28 07:41:28 Dvorkin: you'd need to create two different-named packages - then you can select between those in the image recipe Sep 28 07:49:45 bluelightning, I know. I created and trying to select, but... Sep 28 07:51:02 bluelightning, several other packages are depending on that package. so BB have to decide early what version to install Sep 28 07:53:16 bluelightning, let me explain. I created two packages: rA and rB. they both are prooviding VIRTUALLY the same package "r". some other pacages are depending on virtual package "r". If I specify "rA" or "rB" in the image recipe, it doesn't work. Sep 28 08:33:02 I need advice on ethernet-over-usb, otg, imx7, android. Any good channels for such Qs? Sep 28 08:44:01 Dvorkin: well bitbake doesn't make that decision at that point, it's the package manager Sep 28 08:44:44 I'm unsure as to the relative capabilities of each package manager when it comes to "virtual" dependencies, I know some of them cannot handle it Sep 28 09:04:53 bluelightning, yes. i'm using rpm+dnf and it says, two recipes and one is already installed and can't iverinstall the another. I don't want to install both, but it tryes Sep 28 09:05:10 iverinstal - overinstall Sep 28 09:59:57 Dvorkin: there is no way it will allow you to install both unless you use something like the alternatives system to let you select between files that would otherwise overwrite eachother Sep 28 10:47:30 bluelightning, "alternatives"? oh, that's an idea. I can create the symlinks Sep 28 12:15:16 khem: That's from qemuarm Sep 28 12:15:51 khem: How do you tell GDB that a target is soft or hard float, for arm, is perhaps the issue? Sep 28 17:53:08 * armpit thats weird layerindex does not show portmap Sep 28 19:43:10 armpit: aa4bc52 portmap: remove recipe Sep 28 19:44:06 done after 2.3, before 2.4/rocko though... Sep 28 19:44:50 and to save everyone interested the lookup: It was dropped from Debian years ago, and superseded by rpcbind (which we also ship). Sep 28 19:44:56 vmesons, ah, thanks. but it does not even show up when I change the branch version Sep 28 19:46:20 hmmm, it shows up for me on pyro Sep 28 20:34:37 it looks like having a bad filename in FILESYSTEM_PERMS_TABLES doesn't trigger an error, is that something that should be fixed? Sep 28 20:37:27 what I'm seeing on pyro is silently proceeding w/o a table Sep 28 20:48:58 smurray: that does sound like a bug yes Sep 28 20:50:57 bluelightning: okay, I'll tinker and try to come up with a patch Sep 28 20:51:45 smurray: thanks Sep 28 20:52:47 smurray: hmm, actually it looks like the default table at least contains paths that may or may not exist, so I don't know if we can error Sep 28 20:54:17 bluelightning: I meant when the filename to replace files/fs-perms.txt that's given in FILESYSTEM_PERMS_TABLES doesn't exist Sep 28 20:54:35 smurray: oh right yes, definitely in that case Sep 28 20:58:53 bluelightning: okay **** BEGIN LOGGING AT Thu Sep 28 23:16:10 2017 Sep 28 23:40:30 does a bb.error cause anonymous python to stop from that point on? Sep 28 23:41:21 no Sep 28 23:41:25 that's what fatal is for Sep 28 23:42:12 i'm not really a fan of error(). it shows an ERROR message, makes bitbake return non-zero, but doesn't halt the build. a transient error message that occurs from scratch but not from sstate will then result in inconsistent exit codes depending on whether its from scratch or not Sep 28 23:42:24 * kergoth prefers either warning() or fatal() Sep 28 23:47:03 kergoth: ok so it's the same as print msg; return; Sep 28 23:47:24 or bb.warning(); return Sep 29 00:44:03 mattsm: well it does register as a failure for the build, whereas the other two don't Sep 29 00:45:13 bluelightning what about anon python? Sep 29 00:46:15 actually looking at this: http://git.openembedded.org/openembedded-core/commit/?id=946b6623154e748a0d75ff498802a720aeec27a6 Sep 29 00:46:22 it appears to return, and then things proceed Sep 29 00:46:24 mattsm: should still cause that, if it doesn't that would be a bug... but we have a log handler whose sole job is to listen for logs of level ERROR and that should work everywhere Sep 29 00:46:27 I dont' even see the warrning mesage Sep 29 00:46:35 it won't stop the build in any way though Sep 29 00:46:51 The above patch, does seem to in fact return Sep 29 00:47:30 that surely ought to be bb.fatal() there Sep 29 00:49:17 yea, I was thinking that too... Sep 29 00:49:20 but there are cases Sep 29 00:49:43 when you want to remove an EXTERNALSRC and go back to something else Sep 29 00:50:04 And your assignment doesn't let you do the proper unset so you set to "" Sep 29 00:50:42 it's just checking if it evaluates to true though so if it's "" that should be fine... Sep 29 00:50:54 relatedly, we semi-recently changed LIC_FILES_CHKSUM mismatches to be errors, I think that's wrong Sep 29 00:51:05 (non-fatal errors, I mean) Sep 29 00:52:18 that seems wrong... Sep 29 00:52:48 well i need to sort throug hthis tomorrow Sep 29 00:52:55 can't think striaght ;) Sep 29 01:53:27 mattsm: bb.error just prints a message and sets an exit code, it won't return from the function for you, you'd need fatal to halt there. Sep 29 01:55:36 kergoth where is the code for this bb.error (i'm brain dead)... Sep 29 01:55:50 the issue for me is I can't seem to find the message being printing so I think I'm all messed up Sep 29 01:58:27 bitbake/lib/bb/__init__.py pulls in the bits from bitbake/lib/bb/msg.py **** ENDING LOGGING AT Fri Sep 29 03:00:01 2017