08:32:35 <bero2> #startmeeting
08:32:35 <linarobot> Meeting started Thu Dec 10 08:32:35 2015 UTC.  The chair is bero2. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:32:35 <linarobot> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:32:45 <yspan> hi
08:33:23 <bero2> so it looks like the sprint will be in the week of January 4 and in Shenzhen (to avoid visa trouble)
08:33:52 <bero2> it's not 100% fixed yet, but please try not to make other plans for that week for now
08:34:01 * xruxa-w having problems with IRC this morning
08:34:04 <bero2> #topic daily updates
08:34:29 <bero2> sunao: do you want to go first?
08:34:41 <sunao> What did you do?
08:34:42 <sunao> 1。 building the whole system, but unfortunately, the system can't bootup. many processes failed, even the init. trying to fix this.
08:34:44 <sunao> what are you going to do?
08:34:45 <sunao> 1.  continue porting ptmalloc2.
08:34:47 <sunao> what problems do you have?.
08:34:48 <sunao> 1.  Is there anyway to verify the bionic libc on Host? it is really hard to debug it on board.
08:34:50 <sunao> 2. one hikey board I borrowed from others  was burned and can't bootup. another one has hdmi problem.
08:35:24 <sunao> that's all
08:36:49 <bero2> probably host bionic isn't very meaningful because of all the asm code...
08:37:54 <Qian_> test
08:38:26 <bero2> but should be possible by building for an x86_64 target
08:38:53 <bero2> Qian_: test worked, do you want to go next?
08:39:00 <sunao> thanks bero, I will  have a try
08:39:13 <Qian_> ok
08:39:24 <Qian_> What did you do?
08:39:24 <Qian_> Investigate how to get the application info from ActivityManagerService
08:39:24 <Qian_> Investigate how to get the meminfo(Not in details)
08:39:26 <Qian_> ProcessList.java has defined the MAX_CACHE_APPS = 32, and MAX_TMPTY_TIME,
08:39:28 <Qian_> and MAX_EMPTY_APPS(and many others), seems android has thought about it.
08:39:30 <Qian_> What are you going to do?
08:39:32 <Qian_> Investigate how to force swap the memory of specific process
08:39:34 <Qian_> What problems do you have?
08:39:36 <Qian_> merge libjpeg-turbo to PRO 5, my company colleague found when it opened 20M pictures, it crashed when call the decoder, the probability > 50%, But 16M and 9M seem good . Could you help me examine the case on juno and Hikey borads?
08:40:35 <Qian_> test
08:40:45 <bero2> hmm, probably not just the OOM killer at work?
08:41:21 <Qian_> bero2: I have asked them, they said it memory seems good
08:42:00 <bero2> do you have a sample that crashes it?
08:42:10 <zhizhoutian__> Qian_: why you want to force swapping a specific process's memory?
08:42:15 <Qian_> bero2: Do you have tested the big pictures before on juno and Hikey borads
08:42:43 <Qian_> zhizhoutian__: for save memory
08:43:27 <Qian_> zhizhoutian__: Android has the mechanism now, the MAX_CACHE_APPS is the biggest apps can cached
08:43:27 <zhizhoutian__> Qian_: but maybe we could not do now know when to swapping them
08:44:16 <Qian_> zhizhoutian__: the oldest we don't use it
08:44:29 <liuyq> Qian_, could you share the 20MB pictures you used for test
08:44:30 <zhizhoutian__> Qian_: The inactive one in LRU
08:44:36 <bero2> Qian_: yes, and on Nexus devices too. Not sure if they were that large though
08:44:47 <sunao> Qian_, Does the original libjpeg works well with the 20MB image?
08:45:07 <Qian_> bero2: Could you create a issue , that i will push the logs on it
08:45:49 <Qian_> sunao: it works well with old one
08:46:23 <Qian_> liuyq: Yes, i will share the picture, and share the logs
08:47:40 <zhizhoutian__> Qian_: but the most oldest one is decided by Linux - the inactive anon page in LRU, right? they are not the special process
08:47:57 <Qian_> bero2: Maybe they don't find the sample, it happens when click the thumbnail from camera to open the picture by gallary app
08:49:22 <Qian_> zhizhoutian__: It's truely LRU, you can image there is only 32 cache-buffer for the applications, if 33 apps opens, sorry, the first one will be swap out
08:49:47 <zhizhoutian__> Qian_: oh, i got your mean.
08:50:10 <zhizhoutian__> Qian_: you swap out the oldest app, not page
08:50:19 <zhizhoutian__> Qian_: I see ... interesting.
08:50:30 <Qian_> zhizhoutian__: I think you can use command for test
08:50:55 <bero2> Qian_: sure, will create an issue
08:50:59 <Qian_> zhizhoutian__: you are right
08:51:30 <Qian_> bero2: Thanks, That's all, any new will upload on it
08:51:54 <bero2> xavierhsu: next?
08:52:00 <xavierhsu1> ok
08:52:14 <xavierhsu1> What did you do?The function "jpeg_write_scanlines" can normally be called, but the problem still exist.
08:52:26 <xavierhsu1> What are you going to do?I need to trace another code for finding the error of jpeg.
08:52:41 <xavierhsu1> What problems do you have?Please give me some suggestion.
08:52:55 <xavierhsu1> Bero, I send the mail to you.
08:52:57 <liuyq> Qian_,  please upload the  pictures and logs here: https://bugs.linaro.org/show_bug.cgi?id=1938
08:53:24 <Qian_> liuyq: i will, thanks
08:53:40 <xavierhsu1> Bero, could you give me some suggestion, thanks.
08:54:02 <bero2> I'll take a look... maybe libjpeg uses static global vars or something
08:54:25 <xavierhsu1> Bero, ok. thanks.@@
08:55:20 <bero2> yspan: next?
08:55:38 <yspan> What did you do?
08:55:38 <yspan> Built libpng with --coverage. It also required -fprofile-arcs in LDFLAGS. clang still complained some link errors, and I solved them by switching back to gcc.
08:55:39 <yspan> What are you going to do?
08:55:39 <yspan> Collect coverage data *.gcda to analysis.
08:55:39 <yspan> What problems do you have?
08:55:42 <yspan> no
08:59:56 <liuyq> test
09:00:33 <bero2> liuyq: next?
09:00:48 <liuyq> What did you do?
09:00:49 <liuyq> LAVA HiKey  devices have problem to run jobs
09:00:49 <liuyq> Investigated on LMG-945 about the libc.a size problem.(might caused by TARGET_2ND_CPU_VARIANT)
09:00:49 <liuyq> Revered the SELinux change for CONFIG_MODULES for HiKey
09:00:49 <liuyq> Verified 6.0.1_r3 on local side
09:00:49 <liuyq> What are you going to do?
09:00:53 <liuyq> Continue investigation on  LMG-945
09:00:55 <liuyq> Follow up LCR jobs status
09:00:57 <liuyq> What problems do you have?
09:00:59 <liuyq> None
09:01:07 <liuyq> bero2, the hikey devices on LAVA seems back now
09:01:12 <liuyq> bero2, will follow up the status
09:01:48 <bero2> Good...
09:02:09 <bero2> zhizhoutian__: next?
09:02:19 <zhizhoutian__> What did you do?
09:02:19 <zhizhoutian__> 1. Investigate Linux swap strategy
09:02:19 <zhizhoutian__> 2. Read Linux swap/zramfs code
09:02:19 <zhizhoutian__> What problems do you have?
09:02:19 <zhizhoutian__> Want some document discribe Latest Linux swap code clearly. <<Understanding The Linux kernel 3th>> is too old.
09:04:09 <bero2> probably need to talk to KWG
09:04:42 <bero2> eric___: next?
09:04:59 <zhizhoutian__> bero2: Sorry, but what is KWG?
09:05:05 <eric____> What did you do?
09:05:05 <eric____> rebuild the chromium provided by sprd without system libs, fix some build issues;
09:05:05 <eric____> Discuss with Liuyq for build with system libs, but didn’t get a solution for it;
09:05:05 <eric____> What are you going to do?
09:05:05 <eric____> build the clean code on hackbox;
09:05:06 <eric____> What problems do you have?
09:05:06 <eric____> There’s no Android.mk for chromium, how to build it with system libs? I need Bero to give some advises.
09:05:06 <zhizhoutian__> bero2: a book, or a orig?
09:05:07 <eric____> Can’t login hackbox for ssh key, I have alread update the new key to linaro server, but still permission denied.
09:06:24 <bero2> KWG is the Kernel Working Group, they typically they'll know best how Kernel internals work
09:06:40 <zhizhoutian__> bero2: I see.
09:07:09 <eric____> local chromium codebase can work now, but there's no
09:07:33 <bero2> eric___: I'll take a look at stuff on hackbox
09:08:07 <liuyq> eric____, you can sent mail to its@ with your problem or talk in the #linaro-systems channel
09:08:22 <bero2> eric___: the chromium makefiles are autogenerated by gyp, liuyq has a patch for an older version you may want to look at
09:09:02 <eric____> yes, I know, so i will take some time to investigate on gyp
09:10:06 <eric____> ssh issue maybe not yours
09:10:34 <eric____> because, there're several keys in my disk, I need a way to manage them
09:10:49 <bero2> you may want to check for errors with ssh -v -v
09:11:41 <bero2> wuhai: next?
09:11:51 <wuhai> OK
09:12:04 <wuhai> What did you do?
09:12:04 <wuhai> 1.Try porting EMMC swap function to T-shark baseline
09:12:04 <wuhai> 2.Verify GCC5.3 effect on T-Shark baseline.
09:12:04 <wuhai> what are you going to do?
09:12:04 <wuhai> 1.Verify EMMC swap function effect on T-Shark baseline.
09:12:04 <eric____> liuyq: thanks
09:12:04 <wuhai> what problems do you have?
09:12:04 <wuhai> I will continue LMG-963 task?
09:12:16 <liuyq> bero2, my old work was on .mk file directly, no changes on the gyp files:(
09:14:12 <eric____> liuyq: bero2 yes, what I want to investigate is how the gyp files generate .mk? It shoudl be much better If you can explain to me directly
09:14:41 <bero2> wangjian: next?
09:15:13 <bero2> eric___: I have no idea about gyp, just know what it does... Anyone else?
09:15:43 <wuhai> bero2: i wil continue LMG-963 task ?
09:15:44 <wangjian> What did you do?
09:15:44 <wangjian> Has build the T-shark system and boot by linaro-gcc 5.3,and found that has no effect on reduce the memory usage compare with the version which donot merged linaro patchset.
09:15:44 <wangjian> What are you going to do?
09:15:44 <wangjian> I think I can work on LMG-924: https://projects.linaro.org/browse/LMG-924
09:15:44 <wangjian> What problems do you have?
09:15:44 <wangjian> Does this task can be assigned to me?
09:17:25 <eric____> bero2: OK, I will investigate it
09:18:44 <liuyq> bero2, the test result on kernel with gcc5.3+ android with gcc5.3 + linaro patchset consume 4MB more the the original images without linaro patchset
09:19:14 <bero2> wuhai: yes, unless you have a better idea
09:19:14 <liuyq> bero2, any idea about that, and what's need to try next?
09:20:08 <wuhai> bero2 OK,i will do it
09:20:22 <bero2> wangjian: yes, feel free to take it - I think Xavier will be busy with libjepg for a while
09:20:51 <wangjian> bero2: ok,thanks
09:21:22 <bero2> yuxing__: next?
09:21:47 <bero2> liuyq: that's pretty much what I saw on Nexus 7 as well...
09:22:07 <bero2> liuyq: didn't have the time to investigate yet though
09:22:15 <yuxing__> What did you do?
09:22:16 <yuxing__> 1. Investigate whether the page which be mergable by ksm can be swaped out
09:22:16 <yuxing__> 2. continue test the ksm parameter : pages_to_scan and sleep_millisecs
09:22:16 <yuxing__> What are you going to do?
09:22:16 <yuxing__> 1. continue above work
09:22:16 <yuxing__> What problems do you have?
09:22:16 <yuxing__> 1. How to test the affect of ksm for the whole system?
09:22:17 <yuxing__> 2. whether exist some tools for test ksm ?
09:26:10 <yuxing__> bero: can you help me solve the problem ?
09:27:59 <Qian_> yuxing__: try application "cool tool"
09:29:05 <yuxing__> Qian: thanks a lot
09:30:20 <bero2> yuxing__: you may also want to talk to sumits
09:30:20 <yuxing__> Qian, it is a APK ?
09:30:41 <Qian_> yes, you can google "android cool tool"
09:31:24 <yuxing__> bero2:, Qian: OK thanks
09:35:59 <bero2> #endmeeting