**** BEGIN LOGGING AT Thu Sep 11 02:59:58 2014 Sep 11 07:25:05 good morning Sep 11 07:26:46 hi mckoan Sep 11 08:58:13 morning all Sep 11 08:59:11 bluelightning, morning Sep 11 09:00:53 hi kroon Sep 11 09:05:05 jo Sep 11 09:11:09 hi woglinde Sep 11 09:35:44 do we have a specially patched mkfs which adds the -d option? Sep 11 09:36:17 I'm trying to move away from using sudo and loopback mounts in our in-house build system and was looking to OE for inspiration Sep 11 09:36:49 but I see what we build the sysroot and then pass it into mkfs the with -d option, but I can't find this option documented Sep 11 09:39:22 ah yes ok, I see the patch to add this functionality in e2fsprogs Sep 11 09:40:55 I could be wrong but I think that functionality got merged upstream recently as well Sep 11 09:41:06 yeah I've just seen the status is Backported Sep 11 09:51:34 ok, I see they have been merged but they're not in the latest release yet which is confusing Sep 11 09:52:13 ah, ok... maybe someone should poke them about doing a new release Sep 11 09:52:25 there was a new release recently Sep 11 09:52:30 thats the confusing part Sep 11 09:52:39 maybe they're waiting for a major number release Sep 11 09:52:48 could be Sep 11 09:53:06 I think it would be legitimate to ask when they plan that to occur Sep 11 09:53:08 but the last major number release was 2010 Sep 11 09:53:13 i might pop over to the channel and poke them Sep 11 09:53:22 awesome, thanks Sep 11 10:06:16 Hello All, Can i get url for glibc-2.11.x recipe? Sep 11 10:14:48 bryan: as I've said several times you will find old OE-Classic recipes in our our OE-Classic repository (http://cgit.openembedded.org/cgit.cgi/openembedded/) Sep 11 11:03:26 I could find latest version as glibc-2.10.1 Sep 11 11:03:43 but in eglibc it has version eglibc-2.12 Sep 11 11:04:14 can we use eglibc recipe instaed of glibc Sep 11 11:10:45 if you write / adapt your own recipe you could, yes Sep 11 11:22:22 Hello, I am a new developer for bitbake so please bare with me. I submitted a patch for improving the PR service persistence last month and I have yet to get any reply. Is there anyone that can possibly tell me if it is a mistake on my part? thanks in advance Sep 11 12:12:17 DMT: hi, which mailing list did you send the patch to? I think I recall seeing it but I can't find it at the moment Sep 11 12:13:09 its here: http://lists.openembedded.org/pipermail/bitbake-devel/2014-August/thread.html Sep 11 12:17:16 I should have given the actual post, here it is: http://lists.openembedded.org/pipermail/bitbake-devel/2014-August/004960.html Sep 11 12:17:27 right, just looked at it Sep 11 12:17:53 I have to be honest and say I don't know enough about that section of the code myself Sep 11 12:19:27 RP: ^ Sep 11 12:22:53 bluelightning: Ah, well thanks for your reply. I will stick around, maybe someone else will pop on later that has worked with the PR service. Again, thanks! Sep 11 12:29:56 bluelightning, hangouts often have a limit of 10 people Sep 11 12:30:13 , but you can do a hangout on the air which lets people watch on youtube Sep 11 12:30:17 and saves it after Sep 11 12:30:25 Crofton|work: it is marked as a hangout on air FWIW Sep 11 12:30:39 I have no idea how to make any distinction between participants and viewers though Sep 11 12:30:39 also, we have sqlite in oe-core and meta-guacamayo Sep 11 12:30:59 well, if people hav etrouble getting in, ask people to view via youtube Sep 11 12:31:05 I may do that Sep 11 12:31:26 also, you shuold be able to get lnks to joi nthe hangout and the youtube one once it is started Sep 11 12:31:34 ok, thanks Sep 11 12:31:57 correct, that's how it works Sep 11 12:32:30 hey, it is the local celebrity! Sep 11 12:32:59 LOL Sep 11 12:33:13 we gotta get you a job! Sep 11 12:33:47 _that_ would be appreciated. I do cute embedded management hacks pretty well if someone asks. Sep 11 12:34:12 I can also manage people doing cute embedded nonsense hacks Sep 11 12:36:59 http://www.computerworld.com/article/2605093/laid-off-from-job-man-builds-tweeting-toilet.html Sep 11 12:37:41 DMT: I did see the patch, sorry for not replying Sep 11 12:38:28 DMT: I think my concerns were a) why are we sending sync requests instead of just doing the timeout on the server side? b) can we only sync if the database is dirty? Sep 11 12:38:57 lovely, libdbi in meta-oe and meta-telephony Sep 11 12:45:08 RP: I initially was performing the sync immediately from the daemon thread, however, I did some research on the check_same_thread flag for the sqlite3 connect call. The documentation had no reference, however there were multiple posts about unpredictable exceptions and Python crashes Sep 11 12:45:54 RP: So I decided as a precaution to follow the producer consumer pattern already in place for the syncing as well Sep 11 12:48:13 RP: I am not sure what the problem is with the syncing when the server is dirty. Are you suggesting extending the server so syncing can be done by clients as well? I could add that, I would not take that much effort Sep 11 12:50:49 DMT: I think I misunderstood the patch and a) isn't correct, that does leave b) though Sep 11 12:51:14 DMT: I'm not sure I like calling sync() every 20s Sep 11 12:55:02 DMT: sync is only called when the server is dirty, which is when an operation has been performed by the consumer thread. 20 seconds was just an arbitrary default value I came up with that works with the environment I am working with Sep 11 12:55:22 RP: ^ Sep 11 13:00:26 DMT: how do you know that the operation actually changed data though? Sep 11 13:04:56 RP: In my solution I don't, I was considering making only the specific cases where database activity cause a sync, but I wanted to keep the patch simple so that I could get some feedback from you guys. The non db operations are only ping and getinfo, so adding code for every case seemed a bit overkill Sep 11 13:05:48 DMT: I've looked at this before and its very hard to know if data has changed or not due to the sql query used :/ Sep 11 13:06:10 DMT: we have had significant bottleneck issues caused by sync calls :( Sep 11 13:09:17 RP: aaaaa, its a matter of changes on the db it's self Sep 11 13:09:54 RP: hmmm, I wasn't aware of that, that does complicate things Sep 11 13:12:06 RP: How often are PRs being issued in the cases where there are significant bottlenecks? It would be nice to know what kind of numbers you are dealing with Sep 11 13:16:57 RP: Because, this may be dumb on my part, but wouldn't increasing the timeout time to above average delay between calls to the service reduce the bottleneck enough to make it manageable? Just pointing this out: if another command is sent before the 20second timeout happens the timer is reset and no sync happens Sep 11 13:19:21 RP: But again, I have no idea what beast you are dealing with so it makes most of my argument moot :p Sep 11 13:19:57 DMT: I do remember a ton of hangs, timeouts and a lot of other issues in this area which is why I'm nervous about changing it. I can see the need for what you describe though Sep 11 13:22:32 DMT: Looking at the code again, I think we need a variant of your patch where the code itself maintains a "dirty" flag if an INSERT/UPDATE db call has been used, then we start a timeout to sync things Sep 11 13:23:53 RP: I can work on that, it shouldn't take much time. Should I change the timeout so it is done by the daemon rather than sending it to the consumer? Sep 11 13:24:56 DMT: I think it should be internal to the daemon, yes Sep 11 13:26:49 RP: OK, I will change it back, that will make the code simpler, What about the timeout? I am guessing that will also come from experience, but an estimate for a good number based on your system would be nice Sep 11 13:27:23 DMT: I'd guess around 60s would probably be best Sep 11 13:30:12 RP: OK, that should be doable. Sep 11 13:30:44 RP: I don't think I am missing anything. Thanks a lot for the reply! :) Sep 11 13:31:20 RP: If I get stuck on anything or need some feedback I know where to find you Sep 11 13:59:53 I'm looking for recipes for clang and llvm Sep 11 14:01:08 heeen use the layer search Sep 11 14:01:25 http://layers.openembedded.org/layerindex/branch/master/layers/ Sep 11 14:02:14 woglinde: it doesn't show any results Sep 11 14:02:30 there are some llvm results Sep 11 14:02:35 but not clang Sep 11 14:03:40 some guys have done some work with those? Sep 11 14:05:10 ah ok nevermind Sep 11 14:05:20 I can find llvm3.3 Sep 11 14:05:25 but not clang Sep 11 14:06:48 do these build targeht or host binaries Sep 11 14:09:50 hmm Sep 11 14:10:36 https://lists.yoctoproject.org/pipermail/yocto/2014-June/020358.html Sep 11 14:10:55 this reciĆ¼pe just builds clang from within llvm Sep 11 14:15:57 clang wants to be built from within llvm's source tree, does oe support this in a separate recipe? Sep 11 14:40:35 what does BBCLASSEXTEND do Sep 11 15:26:26 I am trying to build llvm as a host binary, I added inherit native but it still is configured with host=arm Sep 12 00:22:47 heeen: checkout https://github.com/kraj/meta-clang Sep 12 00:22:55 its not final yet Sep 12 00:23:10 but its going to go forward soon **** ENDING LOGGING AT Fri Sep 12 02:59:58 2014