**** BEGIN LOGGING AT Mon Dec 02 02:59:58 2013 Dec 02 07:52:34 morning Dec 02 08:52:06 morning Dec 02 09:05:52 Tofe: morning Dec 02 09:06:14 Tofe: we're really going to have memory problems Dec 02 09:48:11 morphis: ah, it's that bad... Dec 02 09:48:39 I looked a bit two weeks ago, and it seems like there is a lot of big textures allocations Dec 02 09:49:44 I don't know if that may explain the issue; and we do need most of those, because the apps are actually running and drawing in the background Dec 02 09:50:44 I don't know if it's possible of meaningfull to compare with LSM on TP Dec 02 10:01:36 morphis: one problem is that I don't really know any good way to analyse memory consumption at runtime, moreover on target device Dec 02 10:08:49 hm Dec 02 10:11:55 Tofe: one thing we can actually do to reduce the memory used by apps Dec 02 10:12:37 * Tofe is all ears Dec 02 10:13:05 we can avoid apps using the noWindow mode Dec 02 10:13:14 as thats causing another QtWebProcess instance Dec 02 10:13:46 well, they don't need any UI, but they still need to be run, isn't it ? Dec 02 10:33:59 Tofe: yes, they don't need any UI but will use a lot memory Dec 02 10:34:22 so for a noWindow app we have 1xwebapp-launcher and n x QtWebProcess Dec 02 10:34:44 I would like to really have only 1 x webapp-launcher and 1 x QtWebProcess per app Dec 02 10:35:41 ok, I didn't know there was n QtWebProcess instances per app; I don't know webos well enough :p Dec 02 10:36:21 when is a QtWebProcess created ? Dec 02 10:39:50 Tofe: it's webkit2 architecture Dec 02 10:40:00 it has a UIProcess and a WebProcess Dec 02 10:40:10 the UIProcess is the webapp-launcher in our case Dec 02 10:40:24 it spawns the WebProcess for processing the web content on it's own Dec 02 10:40:50 but there's only one web content per app, isn't it ? Dec 02 10:40:50 it simply means one WebProcess per web content page Dec 02 10:41:16 no, in case of a noWindow app there is one page for the noWindow part which spawns more pages Dec 02 10:41:50 for example the mail app uses this Dec 02 10:42:08 it has the nowindow part open to listen for incoming mails to show a notification for example Dec 02 10:42:23 so the noWindow part is simply a "background service" Dec 02 10:42:37 was a pretty common pattern in webOS Dec 02 10:42:38 I understand, but why doesn't it happen for UIProcesses ? Dec 02 10:42:56 that we have multiple UIProcesses per app? Dec 02 10:43:06 no no, multiple WebProceses Dec 02 10:43:25 visible or not, I guess an app can always start new pages Dec 02 10:43:43 thats app to the implementation Dec 02 10:43:55 currently we create a new WebView when the app calls window.open Dec 02 10:44:24 ok; so having only one QtWebProcess is not necessarily noWindow-specific Dec 02 10:45:25 take it like this: when a noWindow app starts it default index.html will be loaded into a page which will never be visible (a "noWindow") Dec 02 10:45:37 it's then up to this page to spawn visible pages Dec 02 10:45:46 and each page gets it's own WebProcess Dec 02 10:47:07 ok Dec 02 10:47:38 then, migrating to having only one QtWebProcess is quite a big refactoring, actually Dec 02 11:18:10 Tofe: the thing we can do is: we forbid noWindow Dec 02 11:18:41 we can just ignore it in the implementation and don't allow applications to use it Dec 02 11:18:51 or add a whitelist for a limit count of apps to use it Dec 02 11:19:26 s/limit/limited/ Dec 02 11:20:05 that way a app can't open a second window if it's not allowed so we always have on webprocess and don't use that much memory per app Dec 02 11:22:38 Tofe: http://bpaste.net/show/b67ks0aIiggtem83Tubq/ Dec 02 11:22:46 thats on the gnexus with one app (preware) launched Dec 02 11:28:05 CONFIG_KSM and CONFIG_ZRAM in the kernel also help Dec 02 11:55:31 morphis: but with Qt and QML, maybe there's a big overhead for the first app Dec 02 11:56:17 you mean the first use app? Dec 02 11:56:31 all have 90% of memory use in shared state, except for QtWebProcess which only has half Dec 02 11:57:08 it would be interesting to see mem consumption with 6 apps Dec 02 12:28:53 yes Dec 02 18:48:01 morphis: webapp-launcher doesn't do so well right now Dec 02 18:49:11 doesn't output anything. No error, no process launched, it's a bit decieving Dec 02 18:52:50 ? Dec 02 18:53:07 it works here Dec 02 18:56:44 ah, ok, maybe it's my install that is a bit shaky Dec 02 18:57:23 I just recompiled webapp-launcher and luna-sysmgr, latest commits, and installed them -- also removed luna-webappmanager package Dec 02 18:57:36 Did I miss a point? Dec 02 19:06:09 hm Dec 02 19:06:35 do you have the changes from morphis/work? Dec 02 19:06:41 for meta-webos-ports? Dec 02 19:06:53 * 608dc1f (HEAD, origin/morphis/work, morphis/work) org.webosports.app.firstuse: bump SRCREV Dec 02 19:06:53 * 57f182e Use webapp-launcher as replacement for luna-webappmanager Dec 02 19:06:53 * a1cce7c luna-sysmgr: bump SRCREV Dec 02 19:06:53 * bc856db qtwebkit: use correct SRCREV after used wrong with the last submission Dec 02 19:06:53 * 50e330e webapp-launcher: introduce recipe for webapp-launcher, created from luna-webappmanager Dec 02 19:43:44 I think Dec 02 19:43:57 apart from first use Dec 02 19:44:35 I checkout your branch, we'll see Dec 02 19:44:42 I'll* **** ENDING LOGGING AT Tue Dec 03 02:59:58 2013