08:31:57 <bero> #startmeeting
08:31:57 <linarobot> Meeting started Wed Nov 25 08:31:57 2015 UTC.  The chair is bero. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:31:57 <linarobot> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:32:04 <bero> hi all
08:32:22 <bero> no big news today - so let's get into the daily reports...
08:32:27 <bero> Qian: do you want to go first?
08:32:35 <Qian> ok
08:32:52 <Qian> What did you do?
08:32:53 <Qian> Seems to find out the root cause of hash map error, it’s caused by ISA
08:32:53 <Qian> Replace all the places isa and ::isa with UNIVERSIA:isa
08:32:54 <Qian> Run bzip2 test successfully with perl-5.22.0 on X86
08:32:56 <Qian> What are you going to do?
08:32:58 <Qian> Test bzip2 demo on Android
08:33:00 <Qian> Work around about files md5 checking that haven’t been fixed
08:33:02 <Qian> Will update current fruit on SPEC server
08:33:04 <Qian> Want to hand over the rest perl-5.22.0 updating work to LMG-SC, after they have finished and I will go on transplanting it to android.
08:33:06 <Qian> Linaro patches about libc performance testing on MEIZU MX5 and PRO5, maybe one day work
08:33:08 <Qian> What problems do you have?
08:33:10 <Qian> none till now.
08:34:07 <bero> nice... so with the hash map error fixed, what problems remain?
08:34:11 <Qian> bero: I'd like handle the rest of perl updating work to bogden
08:34:50 * liuyq is here
08:34:57 <bero> sounds like a good plan if he has the time
08:34:59 <Qian> I haven't tested all of them, and seems some errors of spectools when compile it
08:35:56 <Qian> I have some work around about md5 checking, and i think it's a good idea to update to perl-5.22.0
08:36:17 <Qian> if we want to transplant SPEC on android
08:36:18 <bero> ok, so we can't assume the worst problems are all fixed already...
08:36:27 <bero> I agree about 5.22.0 being a good idea
08:36:52 <Qian> Maybe some interfaces can't work, but I think we can fixed it.
08:37:20 <Qian> I expect the result should be good
08:37:35 <bero> yes, I'm sure we can . just not so sure anout the timeframe
08:38:24 <Qian> bogen said yesterday, he may check my code harder if it can work,
08:38:39 <Qian> and i think it can work, now
08:39:02 <bero> that's good, maybe we don't need to defer it for long after all
08:39:23 <Qian> perl-5.22.0 is supported well on android
08:40:37 <Qian> and perl-5.12.3 seems a little hard to debug, i get error even type one blink one, i don't know why
08:40:52 <Qian> blink line
08:41:34 <Qian> I will spend one day time for MEIZU's work tomorrow
08:41:55 <bero> I think there's no point in bothering with 5.12.3, it's the wrong direction...
08:42:29 <bero> that's ok, let me know if you're having any problems with the bionic patches setc.
08:43:02 <Qian> Yes, if our goal is merge android and x86 both to master, so perl-5.22.0 is the better choose now
08:44:12 <Qian> I have merge it on MX5 and PP5, left performance test work to do
08:44:29 <bero> yes, and it's also far more likely that 5.22.0 will work on aarch64 as well
08:44:35 <Qian> OK, that's all, i will keep contacting with bogden
08:44:47 <bero> 5.12.3 was released years before anyone even thought there would ever be a 64bit arm CPU
08:44:56 <bero> so let's do it the right way
08:45:10 <bero> xavierhsu: do you want to go next?
08:45:43 <xavierhsu> ok
08:46:00 <xavierhsu> What did you do?  1. I modify the source code of libjpeg.
08:46:15 <xavierhsu> What are you going to do?   1. I need to investigate that which part belong to encode or decode, because I use android app to read the picture that the system will call libjpeg-enc.so.
08:46:31 <xavierhsu> What problems do you have?  1. The android system will be crashed by calling libjpeg-enc.so.
08:47:37 <bero> any idea what may be causing it already?
08:48:03 <xavierhsu> error message 01-01 00:00:56.311  2576  2591 D JPEGENC_STUB: load: loading libjpeg-enc.so
08:48:03 <xavierhsu> 01-01 00:00:56.311  2576  2591 D JPEGENC_STUB: load: dynamic loading symbol jpeg_CreateCompress
08:48:03 <xavierhsu> 01-01 00:00:56.319  2576  2591 D JPEGENC_STUB: load: loading libjpeg-enc.so
08:48:03 <xavierhsu> 01-01 00:00:56.319  2576  2591 D JPEGENC_STUB: load: dynamic loading symbol jpeg_set_defaults
08:48:03 <xavierhsu> 01-01 00:00:56.324  2576  2591 D JPEGENC_STUB: load: loading libjpeg-enc.so
08:48:03 <xavierhsu> 01-01 00:00:56.324  2576  2591 D JPEGENC_STUB: load: dynamic loading symbol jpeg_set_quality
08:48:03 <xavierhsu> 01-01 00:00:56.326  2576  2591 D JPEGENC_STUB: load: loading libjpeg-enc.so
08:48:04 <xavierhsu> 01-01 00:00:56.326  2576  2591 D JPEGENC_STUB: load: dynamic loading symbol jpeg_start_compress
08:48:04 <xavierhsu> 01-01 00:00:56.328  2576  2591 D JPEGENC_STUB: load: loading libjpeg-enc.so
08:48:05 <xavierhsu> 01-01 00:00:56.328  2576  2591 D JPEGENC_STUB: load: dynamic loading symbol jpeg_write_scanlines
08:48:05 <xavierhsu> 01-01 00:00:56.329  2576  2591 D JPEGENC_STUB: load: loading libjpeg-enc.so
08:48:06 <xavierhsu> 01-01 00:00:56.329  2576  2591 D JPEGENC_STUB: load: dynamic loading symbol jpeg_write_scanlines
08:49:33 <xavierhsu> bero, I not sure, but I need to find whitch part belong to encode.
08:50:01 <xavierhsu> Because I read the picture the libjpeg-enc.so that will be called.
08:51:55 <bero> guess we'll need to trace what exactly is being called (and by what)
08:51:58 <xavierhsu> Bero, I use the original jpeg-turbo (Modify by YongQin), it can be normally be used.
08:52:22 <bero> it's possible that the java interface to libjpeg just initializes the encoder unconditionally or so
08:53:13 <xavierhsu> Bero, yes. Maybe. I need to trace the code.
08:53:27 <bero> probably that's the only way to find out what's happening.
08:53:34 <bero> yspan: do you want to go next?
08:53:44 <xavierhsu> Bero, I will.
08:53:50 <xavierhsu> Bero, Thanks.
08:53:51 <Qian> I have a demo about call libjpeg.so from java
08:53:57 <yspan> What did you do?
08:53:58 <yspan> Tried enable gcov (--coverage) for libpng, but encounter some link errors:
08:53:58 <yspan> error: undefined reference to 'llvm_gcov_init'
08:53:58 <yspan> error: undefined reference to 'llvm_gcda_emit_function' �.etc
08:53:58 <yspan> What are you going to do?
08:53:58 <yspan> Maybe it requires some additional libs -lwhat? I will google it. Or I will try to switch to gcc.
08:54:02 <yspan> Also refactor libpng patch according to liuyq�s comments
08:54:04 <yspan> What problems do you have?
08:54:14 <yspan> none
08:54:47 <xavierhsu> Qian, really could you send the demo file to me. thanks
08:55:18 <xavierhsu> Qian, xavier.hsu@linaro.org
08:55:29 <xavierhsu> Qian, thanks.
08:56:08 <Qian> xavierhsu: OK, i will send it to you later
08:57:09 <bero> yspan: I think you have to pass -fprofile-arcs to the linker (LDFLAGS) for llvm_gcov_init and friends
08:57:19 <bero> liuyq: do you want to go next?
08:57:32 <liuyq> What did you do?
08:57:33 <liuyq> Created draft version of release notes
08:57:33 <liuyq> Investigated about adb hang problem for Hikey on LAVA
08:57:33 <liuyq> Created draft version HOWTOs for Hikey
08:57:33 <liuyq> What are you going to do?
08:57:33 <liuyq> Create draft version release notes for HiKey Board
08:57:35 <liuyq> Continue the investigation about adb hang problem for HiKey on LAVA
08:57:39 <liuyq> Send mail to Fathi/Koen for release notes check
08:57:41 <liuyq> What problems do you have?
08:57:42 <yspan> bero, ok, thanks, that saves my time :p
08:57:43 <liuyq> None
08:57:56 <liuyq> bero, mainly things related to release
08:58:04 <bero> good...
08:58:11 <bero> zhizhoutian__: do you want to go next?
08:58:26 <zhizhoutian__> What did you do?
08:58:26 <zhizhoutian__> - Investigate Sprd memory monitor framework
08:58:26 <zhizhoutian__> run in kernel, add many notifiers into subsystem(zram/ion and so on)
08:58:26 <zhizhoutian__> offer /proc interfaces and logged in dmesg, through those we can:
08:58:26 <zhizhoutian__> watch every process’s memory cost status
08:58:26 <zhizhoutian__> watch every module(zram/ION/mali/task…)’s memory status and lost ration
08:58:26 <zhizhoutian__> ...
08:58:27 <zhizhoutian__> - Learn ASan tools
08:58:27 <zhizhoutian__> used for framework, for checking access memory out of band, especially stack/heap out of band.
08:58:28 <zhizhoutian__> add a “Shadow” zone in reserved virtual memory used for allocating stub variables among program. Every virtual memory access will callback isPoison() and check whether it is illegal
08:58:28 <zhizhoutian__> if we compile it with zygote, then we can watch all Android processes.
08:58:29 <zhizhoutian__> - resolved TShark’s deep sleep issue
08:59:01 <Qian> zhizhoutian__: Could you share the tools with us?
08:59:31 <bero> good...
08:59:53 <zhizhoutian__> Qian: Asan is from google
09:00:00 <zhizhoutian__> Qian: you can download it from github
09:00:18 <zhizhoutian__> Qian: it is very very useful to debug memory issue.
09:00:20 <Qian> zhizhoutian__: the tools you mentioned yesterday
09:00:58 <bero> brb, got to switch to mobile for a second, I have a dog going crazy, have to let him go out
09:01:00 <Qian> zhizhoutian__: writen by lingming, yes?
09:01:18 <zhizhoutian__> Qian: sure, you can rejoin Spreadtrum or Linaro
09:01:34 <zhizhoutian__> Qian: then you will get those tools:)
09:01:46 <Qian> zhizhoutian__: haha
09:02:23 <zhizhoutian__> Qian: :)
09:02:48 <bero2> OK back...
09:03:25 <bero2> eric___: do you want to go next?
09:03:31 <eric____> fine
09:03:46 <eric____> Try to make subdirectory with scan-build, but find scan-build doesn’t support mm, mmm commands when compiling;
09:03:56 <eric____> Log the compile detail when compiling the whole android, but found no clue in the log file;
09:04:34 <eric____> now, I'm trying to combine all these commands to a script, and wait for reslut
09:04:47 <eric____> What are you going to do?
09:04:54 <eric____> Try to make scan-build support mm, and mmm;
09:05:01 <eric____> What problems do you have?
09:05:15 <eric____> Scan build can’t work with mm and mmm, so it’s hard to make the subdirectory, try to find a way;
09:05:52 <bero> sounds painful
09:06:11 <bero> wuhai: do you want to go next?
09:06:31 <wuhai> OK
09:06:34 <eric____> now, it becomes, scan-build bash -c "./test.sh"
09:06:48 <wuhai> What did you do?
09:06:48 <wuhai> 1.Sync Tshark code, merge changes to the baseline, compile, brush machine, test the modified content, look normal.
09:06:48 <wuhai> 2.Analysis LocationManagerService architecture code, understand the workflow of the service
09:06:48 <wuhai> What are you going to do?
09:06:48 <wuhai> 1.Continue to analyze the structure of code LocationManagerService
09:06:49 <wuhai> What problems do you have?
09:06:49 <wuhai> None
09:06:53 <eric____> yes, completely mad, ;)
09:07:53 <bero> wangjian: do you want to go next?
09:08:21 <wangjian> What did you do?
09:08:22 <wangjian> Start to analyze the launcher code, found launcher memory primarily used in resource load, such as layout, string, picture, bitmap, etc. and another part of lots of memory usage is Instance the java class.
09:08:22 <wangjian> What are you going to do?
09:08:22 <wangjian> Check whether it can reduce unnecessary resource loading to reduce memory usage.
09:08:22 <wangjian> What problems do you have?
09:08:22 <wangjian> There are too many resource to check.
09:09:22 <Qian> wangjian: resourse manager ?
09:09:26 <bero> yes, probably hard to get some control over when resources are loaded or unloaded...
09:10:00 <wangjian> Qian: no ,launcher resourse
09:10:13 <bero> sunao: do you want to go next?
09:10:33 <sunao> What did you do?
09:10:34 <sunao> 1. Read  the code of dlmalloc, to see if we can save more memory.
09:10:36 <sunao> 2. Some local task.
09:10:37 <sunao> what are you going to do?
09:10:39 <sunao> 1.  Continue read the code of dlmalloc, Not sure whether we can save more memory
09:10:40 <sunao> what problems do you have?.
09:10:42 <sunao> 1.  None.
09:11:07 <wangjian> bero: I think so
09:11:14 <sunao> that's all from my side
09:12:01 <bero2> ok
09:12:15 <bero2> Luther: do you want to go next?
09:14:42 <zhizhoutian__> bero2: as i know luther is busying with training;)
09:15:47 <bero2> ah ok
09:15:59 <luther> ok
09:16:09 <luther> sorry for later response
09:16:16 * luther What did you do?
09:16:16 <luther> 1. Worked on company's urgent tasks
09:16:16 * luther What are you going to do?
09:16:16 <luther> 1. Work on company's urgent tasks
09:16:16 * luther What problems do you have?
09:16:17 <luther> 1. So many company's tasks are blocking there
09:17:30 <bero2> OK...
09:17:51 <zhizhoutian__> bero2: as you said yesterday, next step is figuring out the optimal size
09:18:07 <zhizhoutian__> bero2: do you have any advance advise?
09:18:37 <zhizhoutian__> bero2: I mean the zram optimal things
09:18:48 <bero2> zhizhoutian__: no idea... we need to find the point where it saves most memory without slowing down things too much
09:19:59 <zhizhoutian__> bero2: maybe i need to have most acknowledge of zram fist.
09:20:40 <bero2> yes
09:20:42 <Qian> zhizhoutian__: Do you have tested the performance and energy cost?
09:20:47 <bero2> can't hurt...
09:20:56 <zhizhoutian__> bero2: there should be some tools to measure speed/memory save
09:21:23 <zhizhoutian__> Qian: no, i am a fool on power
09:21:41 <bero2> Maybe sumits (kernel team) knows more
09:22:17 <zhizhoutian__> bero2: i see...
09:22:35 <zhizhoutian__> bero2: can i ask him directly?
09:22:55 <zhizhoutian__> bero2: will he ignore me?(possible ;)
09:23:59 <bero2> zhizhoutian__: sure, he won't ignore you
09:24:11 <Qian> zhizhoutian__: if it works worthy, after you uploaded, Our company team can do it
09:24:21 <bero2> he's in India, so timezone should be ok
09:24:43 <zhizhoutian__> bero2: ok, thanks, i will try to ask some help;)
09:25:31 <zhizhoutian__> Qian: No problem
09:26:48 <Qian> zhizhoutian__: I think you can see the different  of CPU cost and pay attention to waking up from sleep
09:27:24 <zhizhoutian__> Qian: I think Meizu can use sprd chips directly ;)
09:28:00 <Qian> zhizhoutian__: haha, sounds a good idea :)
09:28:17 <zhizhoutian__> Qian: hah;)
10:18:54 <bero> #endmeeting