Joe's Notepad

CIFAR load and save to picture

2019-04-29T02:55:00.000-07:00

import pickle as p
import numpy as np
from PIL import Image
import os

def unpickle(file):
with open(file, 'rb') as fo:
dict = p.load(fo, encoding='bytes')
return dict

'''
batch file key:
b'batch_label'
b'labels'
b'data'
b'filenames'
'''
CIFAR10PATH = "E:\\project\\python\\cifar10\\cifar-10-batches-py\\"
DATABATCH = "data_batch_"
OUTPUTFOLDER = ".\\images\\"
#E:\project\python\cifar10\cifar-10-batches-py
if __name__ == "__main__":
for i in range(1,2):
batchName = CIFAR10PATH + DATABATCH+str(i)
print("batch name:"+str(batchName))
cifar = unpickle(batchName)
for d in range(10000):
labelNum = cifar[b'labels'][d]
imgData = cifar[b'data'][d]
#remove utf-8 name with b'
filenames = cifar[b'filenames'][d].decode("utf-8")

#If the dimension with length 3 is the RGB axis, then you want it at the end.
#If your array has shape (3, R, C), then np.transpose(img, (1, 2, 0))
#will have shape (R, C, 3).
imgData = imgData.reshape(3, 32, 32) #CIFAR default format 3 channels with RxC
#(0, 1, 2) no change (3, R, C) ==> transpose to (1, 2, 0) => (R, C, 3)
imgData = imgData.transpose(1, 2, 0)
img = Image.fromarray(imgData, 'RGB')
folderName = OUTPUTFOLDER + str(labelNum)
if not os.path.exists(folderName):
os.makedirs(folderName)
img.save(folderName+"\\"+str(filenames).strip()+'.jpg')

android trace

2017-11-15T08:46:00.000-08:00

traceView -> for app

systrace ->focus on kernel ftrace(surfaceflinger, )
https://kknews.cc/zh-tw/news/enbv3on.html
kernel -> ftrace, app-> trace class
/tools/systrace.py

追到app問題後，也要使用traceView來協助分析

mt.exe 停止運作 MSB6006

2017-09-27T07:26:00.001-07:00

vs2010

Property -> linker -> Manifest File -> Generate Manifest -> No(Manifest:No)

arduino avrdude upload hex file by command

2017-09-20T10:05:00.000-07:00

Arduino Uno

avrdude -c arduino -P com4 -b 115200 -p ATMEGA328p -U flash:w:camBox.ino.hex
or
avrdude -c arduino -P com4 -b 115200 -p m328p -U flash:w:camBox.ino.hex
or
avrdude -c arduino -P com4 -b 115200 -p ATMEGA328p -U flash:w:E:\camBox.ino.hex:i

Arduino Mega

avrdude -c wiring -P com4 -b 115200 -p m2560 -U flash:w:camBox.ino.hex

avrdude -c wiring -P com4 -b 115200 -p ATmega2560 -U flash:w:E:\camBox.ino.hex:i

partno list
http://www.nongnu.org/avrdude/user-manual/avrdude_4.html

or see the avrdude.conf

步進馬達

2016-12-29T18:14:00.001-08:00

幾相幾線?

相:最直接影響的是精度，越多相，精度越細
兩相: 激磁後，1.8DEG

線:最直接影響就是unipolar or bi-polar, 單電流激磁，按照順序即可順轉反轉
雙電流則，電留有方向性來控制正反轉

驅動IC:ULN2003，7組達靈頓電路，可以將每組並聯使用，放大驅動電流來達到驅動基本電流的需求

adb lauch each app from android

2016-02-18T00:31:00.002-08:00

adb shell pm list package -f
adb shell pm list package #list all package

launch each package by monkey

adb shell monkey -p ${package_name} -c android.intent.category.LAUNCHER 1;

github usage note

2016-02-08T07:31:00.000-08:00

github 有個情境

假設要共同開發一個project ex: http://github.com/dev.git
這時你可以fork 到你的github, 經過開發時間透過pull request，跟夥伴貢獻了不少commit

這時該如何更新最新的version到你fork的git?

first, 先建立一個upstream, 可任意命名，就稱upstream吧

git remote add upstream http://github.com/dev.git

then, fetch remote, git remote -v --可以看origin 跟 remote的 git

git fetch upstream
git checkout master

這時可以更新你的git by rebase
git rebase upstream/master
也可透過merge方式來更新git
git merge upstream/master

rebase push 到你的git
git push -f origin master

ref:
http://stackoverflow.com/questions/7244321/how-to-update-a-github-forked-repository

開發local branch 一陣子後 commit 並merge 成 master
dev
master
git checkout dev
git commit -a
git checkout master
git merge dev

更新upstream 後，merge 到 local dev branch
git checkout master
git fetch upstream
git merge upstream/master
==> commit to your fork
git checkout dev
git merge master

C structure on 32bits/64bits compiler

2015-02-26T09:56:00.000-08:00

64bits C structure alignment = 8
32Bits C structure alignment = 4

if u have a data which is building by 32Bits CPU and parsing by 64Bits for structure raw data. it is going to face the different compiler issue.

although the data structure can be fixed the compliler pre-process as #pragma pack(1) to avoid this issue but it is not useful on the complex data structure. In this case, sending of data on different OS must flow the popular data format to avoid the structure raw data problem.

[XML Parser]building expat example by static prebuilt library

2013-09-26T08:50:00.002-07:00

expat , free xml parser developed by C, which was made by James Clark.

we can download the free source by CVS as below link:

cvs -z3 -d:ext:developername@expat.cvs.sourceforge.net:/cvsroot/expat co modulename

you can download the pre-build win32 library by expat website.

using VC10 to build expact example by static library. remember that XML_STATIC you must pre-define.

the static library is libexpatMT.lib

build android adb for windows

2013-09-05T00:07:00.002-07:00

http://source.android.com/source/downloading.html

note that checkout specific branch will make you downloading fast.

I used below command

$ repo init -u https://android.googlesource.com/platform/manifest -b android-4.2.2_r1
$ repo -j10 sync

download complete, u should compiler the adb for Linux first

make -j4 acp adb showcommands

finish building Linux adb and then u can build for windows.

make -j4 USE_MINGW=y adb

Install: out/host/windows-x86/bin/AdbWinUsbApi.dll
cp -fp out/host/windows-x86/obj/EXECUTABLES/AdbWinUsbApi_intermediates/AdbWinUsbApi.dll out/host/windows-x86/bin/AdbWinUsbApi.dll
host Prebuilt: AdbWinApi (out/host/windows-x86/obj/EXECUTABLES/AdbWinApi_intermediates/AdbWinApi.dll)
out/host/linux-x86/bin/acp -fp development/host/windows/prebuilt/usb/AdbWinApi.dll out/host/windows-x86/obj/EXECUTABLES/AdbWinApi_intermediates/AdbWinApi.dll
Install: out/host/windows-x86/bin/AdbWinApi.dll
cp -fp out/host/windows-x86/obj/EXECUTABLES/AdbWinApi_intermediates/AdbWinApi.dll out/host/windows-x86/bin/AdbWinApi.dll
Install: out/host/windows-x86/bin/adb.exe
cp -fp out/host/windows-x86/obj/EXECUTABLES/adb_intermediates/adb.exe out/host/windows-x86/bin/adb.exe

Yes!! complete!

Ubuntu 12.04 + android adb

2013-08-08T23:49:00.001-07:00

1st Step:
Download SDK Tools only from http://developer.android.com/sdk/index.html.
My download link is http://dl.google.com/android/android-sdk_r22.0.5-linux.tgz

2nd Step:
untar the SDK tool in your folder and execute the android from $yourPath/android-sdk-linux/tools/
if you didn't have install the jdk, please install by the prompt. sudo apt-get install openjdk-7-jdk.

3rd Step:
install the Android SDK Tools, Android SDK Build-tools by the Android SDK Manager as the "
android" execution.

4th Step:
./adb from the $yourPath/android-sdk-linux/platform-tools
Maybe it will be occurred "No such file or directory when you execute. Please install by this prompt -> sudo apt-get install lib32ncurses5 lib32stdc++6

let VB using sharefolder

2013-08-07T19:10:00.003-07:00

sudo usermod -G vboxsf -a [username]

ubuntu 12.04 samba setting

2013-07-18T17:08:00.000-07:00

# sudo apt-get install samba
# sudo apt-get install system-config-samba

DASH --> samba

setting complete~!

pthread condition waiting mutex

2013-01-06T22:47:00.002-08:00

pthread_cond_wait()

再使用此function時，要特別注意，signal 的timing，不然很容易等不到訊號，必須配合mutex

假設有 call function "command" 與 thread A process command 這種狀況。

command()
{
enqueue_command();
signal
}

thread_A()
{
while(1)
{
waiting_signal
while(dequeue());
}
}

如果沒有mutex 輔助的話，可能會發生還沒有waiting 時，signal 不斷的近來，這樣就算queue了之前的command 也無用，因為要等下一次正確時間的先waiting 後 signal 的狀況才能夠一次把這些dequeue乾淨，邏輯是不對的

其實用select 的方式作 sleep 就可以解決這樣的問題，反正指令都已經queue起來了，所以有時後在這種狀況，使用waiting condition 配合mutex 是有點多此一舉。

vim structure 自動完成

2012-11-29T22:21:00.004-08:00

這是一大福音，vim7 已經support 可以像windows 一樣寫code 幫你尋找結構中的成員，不是以前的 ctrl + n的功能了，只要在vimrc 底下，加入以下設定，搞定。 f8 為此功能，f7 就像以前ctrl+n

filetype plugin indent on
set completeopt=longest,menu
imap <f8> <C-X><C-O>
imap <f7> <C-X><C-I>

xargs 好文章

2012-11-20T18:25:00.004-08:00

xargs

xargs用來處理一些輸出結果要來當另一個輸入的arguments碰到的問題。例如

      # find /usr/include/ -name "*.h" | xargs -n 2 diff

-n2是指定有兩兩當成輸出變成diff的argumnets

      # find /usr/include/ -name "*.h" | xargs grep '#ifdef'

正常內定是輸出的一拖拉窟的結果，有用xargs時是一個一個餵給後面的命令

      # find /usr/include/ -name "*.h" | xargs -i cp {} ~/include/

-i 與 { }可以把find的輸出的每一個當成cp的第一個argument， /.就可以當成第二個argument。其實find裡面有-exec這個選項後面也可以用{ }表示一個一個餵給後面程式，而不是一拖拉庫的餵給後面程式。

      # find /usr/include/ -name "*.h" -exec cp {} ~/include/

文章出自http://www.study-area.org/cyril/scripts/scripts/node30.html

defined but not used

2012-04-26T01:28:00.001-07:00

Just saw this thread while searching for solutions to this problem. I post here for completeness the solution I found...

The GCC compiler flags that control unused warnings are

-Wunused-function-Wunused-label-Wunused-parameter-Wunused-value-Wunused (=all of the above)

Each of these has a corresponding negative form with "no-" inserted after the W which turns off the warning (in case it was turned on by -Wall, for example). Thus, in your case you should use

-Wno-unused-function

Of course this works for the whole code, not just compile-time asserts. For function-specific behaviour, have a look at Function attributes.

macbook pro virtualbox + ubuntu 10.04 vbox resolution

2012-04-11T07:39:00.002-07:00

sudo gedit /etc/X11/xorg.conf

http://ubuntuforums.org/showthread.php?t=719126

# xorg.conf (X.Org X Window System server configuration file) # # This file was generated by failsafeDexconf, using # values from the debconf database and some overrides to use vesa mode. # # You should use dexconf or another such tool for creating a "real" xorg.conf # For example: # sudo dpkg-reconfigure -phigh xserver-xorg Section "InputDevice" Identifier "Generic Keyboard" Driver "kbd" Option "XkbRules" "xorg" Option "XkbModel" "pc105" Option "XkbLayout" "us" EndSection Section "InputDevice" Identifier "Configured Mouse" Driver "vboxmouse" Option "CorePointer" EndSection Section "Device" Identifier "Configured Video Device" Boardname "vesa" Busid "PCI:0:2:0" Driver "vboxvideo" Screen 0 EndSection Section "Monitor" Identifier "Configured Monitor" Vendorname "Generic LCD Display" Modelname "LCD Panel 1440x900" Horizsync 31.5-56.0 Vertrefresh 56.0 - 65.0 modeline "800x600@56" 36.0 800 824 896 1024 600 601 603 625 +hsync +vsync modeline "800x600@60" 40.0 800 840 968 1056 600 601 605 628 +hsync +vsync modeline "1280x768@60" 80.14 1280 1344 1480 1680 768 769 772 795 -hsync +vsync modeline "1280x720@60" 74.48 1280 1336 1472 1664 720 721 724 746 -hsync +vsync modeline "1280x800@60" 83.46 1280 1344 1480 1680 800 801 804 828 -hsync +vsync modeline "1440x900@60" 106.47 1440 1520 1672 1904 900 901 904 932 -hsync +vsync Gamma 1.0 EndSection Section "Screen" Identifier "Default Screen" Device "Configured Video Device" Monitor "Configured Monitor" Defaultdepth 24 SubSection "Display" Depth 24 Virtual 1440 900 Modes "1440x900@60" "1280x800@60" "1280x720@60" "1280x768@60" "800x600@60" "800x600@56" EndSubSection EndSection Section "ServerLayout" Identifier "Default Layout" screen 0 "Default Screen" 0 0 EndSection Section "Module" Load "glx" Load "GLcore" Load "v4l" EndSection Section "ServerFlags" EndSection

ubuntu 11.04 tftp install

2011-09-11T09:08:00.000-07:00

joe@ubuntu: sudo apt-get install xinetd tftpd tftp

#vi /etc/xinetd.d/tftp

service tftp

{

protocol = udp

port = 69

socket_type = dgram

wait = yes

user = nobody

server = /usr/sbin/in.tftpd

server_args = /tftpboot

disable = no

}

#sudo /etc/init.d/xinetd restart

vim + ctags + cscope + taglist + Trinity + NERD_tree + Source Explorer

2011-09-11T00:25:00.000-07:00

VIM也用了將近一年半，以往都透過vim + ctags + cscope + taglist開發程式相當習慣，雖在某些功能部份稍嫌不足，但不足以掛齒，最近因為更新VIM，突然忘記vim plug-in 的路徑，請求Google大神後，無意發現有人整理vim的plug-in讓用起來很類似source insight, 似乎操作爽乎。所以乾脆就順便更新一下久沒動過的blogger，讓這些好用的外掛謹記在心。

先下載所有的plug-in

在ubuntu 11.04下安裝流程

先安裝ctags cscope

#sudo apt-get install cscope ctags

Create the folder in ~/.vim/plugin/

將cscope_map.vim 複製到以上資料夾

到你要開發的根目錄下，建立ctags and cscope

#ctags -R

#ctags -R --exclude Makefile, .svn and so on.

#cscope -Rbq

基本上就這時就有vim + ctags + cscope的功能了。

接著把所有plugin *.vim copy 到 /usr/share/vim/vim73/plugin/

這時的外掛就已經安裝完了。

這時只需要修改~/.vimrc 成為你想要的行為。

以下為我的設定方式。一切搞定就緒！！

"ser vim

set number "display number of line

set ai "auto alignment

syntax on "color text

set tabstop=4 "tab is 4 spaces

set shiftwidth=4

set softtabstop=4

set nobackup "no vi backup

set cursorline "display the line for cursor

set hlsearch

set tags=tags;

set mouse=nv

"set autochdir

hi ModeMsg ctermbg=none ctermfg=brown cterm=none

if has("autocmd")

autocmd BufRead *.txt set tw=78

autocmd BufReadPost *

\ if line("'\"") > 0 && line ("'\"") <= line("$") |

\ exe "normal g`\"" |

\ endif

endif

"set taglist

"f12 is the fast link for creating TlistToggle

nnoremap <f12> :TrinityToggleAll

let Tlist_Exit_OnlyWindow = 1

let Tlist_Show_One_File = 1

nnoremap <f7> :TrinityToggleTagList

nnoremap <f8> :TrinityToggleSourceExplorer

nnoremap <f9> :TrinityToggleNERDTree

P.S 我透過更新的方式安裝了11.04，vim 照上面BufReadPost設定還是不會回復到上一次編輯的位置。這時可以透過修改/etc/vim/vimrc BufReadPost的部份，將註解拿掉。

reference:

http://easy72.blogspot.com/2009/05/vim-trinity-tabbar.html

http://stenlyho.blogspot.com/2010/03/vim-trinity-source-explorer-tag-list.html

http://jimmynuts.blogspot.com/2010/12/vim.html

QT 好文

2011-07-25T17:01:00.000-07:00

http://thornbird-life.blogbus.com/logs/115983233.html

android porting note

2011-04-02T07:32:00.000-07:00

kernel porting

s3c6410

android kernel-2.6.32

machine ID error

Uncompressing Linux.........................................................................................

Error: unrecognized/unsupported machine ID (r1 = 0x000009d8).

Available machine support:

ID (hex) NAME

0000065a SMDK6410

Please check your kernel config and/or bootloader.

Refer linux-kernel2.6.32/include/asm/mach-types.h

#define MACH_TYPE_SMDK6410 1626 ==> 0x65A

u-boot wasMINI6410 # bdinfo

arch_number = 0x000009D8

env_t = 0x00000000

boot_params = 0x50000100

DRAM bank = 0x00000000

-> start = 0x50000000

-> size = 0x10000000

ethaddr = 08:90:90:90:90:90

ip_addr = 192.168.2.230

baudrate = 115200 bps

為了麻煩不改變u-boot machine ID 直接改kernel include/asm/mach-type.h，將smdk6410 與 mini6410 的板號直接對調 rebuild 即可。

P.S 有些u-boot 可以直接餵 machine id 在 u-boot 的環境變數即可。mini6410 應該是沒有support。

File systems->Miscellaneous->yaffs2*.

drivers/misc/pmem.c:441: error: conflicting types for 'phys_mem_access_prot'

/home/joe/project/mini6410/android/diff/android_pure_kernel/arch/arm/include/asm/pgtable.h:322: note: previous declaration of 'phys_mem_access_prot' was here

他要call 的是static 的function

這只是名稱有重複使用跟宣告了，把phys_mem_access_prot改個名字就可以了！我給他改成pmem_access_prot

build kernel 看過程內容 V=1

arm-linux-ld: no machine record defined

因為沒完整移植mach-部份，倒置config 沒選擇正確的CPU會出現的錯誤

android 2.2 building

編譯完成後，所有的檔案會再 out/target/product/generic 底下，會有3個產物 ramdisk.img, system.img, userdata.img, 這3個檔案都有對應的目錄，將其所有內容複製到同一資料夾再給予開機格式掛載即可。

注意原始建立出來的 dev 目錄底下是沒有任何node的必須至少建立一個console

sudo mknode console c 5 1

peripheral note

2011-03-29T22:28:00.000-07:00

nand flash with MTD based on linux

--> ioctl-> linux-kernel/mtd/mtdchar.c ==> mtd_ioctl

mtd-abi.h

MEMGETINFO

MEMERASE

MEMGETREGIONCOUNT

git

2011-03-21T06:15:00.000-07:00

git 真的是一個很強大的分散式版本控制器，用途跟SVN目的有點類似但是功能跟使用方法算是天壤之別。

假如今天我要下載samsung linux 的kernel 在網站看到 git://git.kernel.org/pub/scm/linux/kernel/git/kki_ap/linux-2.6-samsung.git

所以可以先將其整個clone到你的local端，

$git clone git://git.kernel.org/pub/scm/linux/kernel/git/kki_ap/linux-2.6-samsung.git

這時候他會將全部的歷史資料download到你的.git裡，下載完成後，可以看到資料夾中是/linux-2.6-samsung.git中kernel最後的版本，這時可以透過branch的方式來看有多少remote的分支版本

$git branch -a

2.6.29

2.6.32

*master

remotes/origin/2.6.24-samsung

remotes/origin/2.6.28-samsung

remotes/origin/2.6.29-samsung

remotes/origin/2.6.31-samsung

remotes/origin/2.6.32-samsung

remotes/origin/2.6.35-samsung

remotes/origin/2.6.36-samsung

remotes/origin/HEAD -> origin/master

remotes/origin/master

這時你可以看到remote之後的分支，假設我們現在再主分支master，為此git kernel最新版本2.6.36-samsung，這時想用2.6.24的版本，就透過checkout方式來切換，這地方有點不習慣，因為SVN通常都是一個個資料夾來存取一個版本的方式，git再這邊是多個版本共用一個資料夾，透過checkout的方式可以切換不同版本。

#可以透過tab鍵幫你自動補齊檔案名稱唷~很方便

$git checkout orgin/2.6.24-samsung

Checking out files: 100% (29816/29816), done.

Note: checking out 'origin/2.6.24-samsung'.

You are in 'detached HEAD' state. You can look around, make experimental

changes and commit them, and you can discard any commits you make in this

state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may

do so (now or later) by using -b with the checkout command again. Example:

git checkout -b new_branch_name

HEAD is now at b30234a... Official Release(09.08.26)

$git branch -a

* (no branch)

2.6.29

2.6.32

2.6.36

master

remotes/origin/2.6.24-samsung

remotes/origin/2.6.28-samsung

remotes/origin/2.6.29-samsung

remotes/origin/2.6.31-samsung

remotes/origin/2.6.32-samsung

remotes/origin/2.6.35-samsung

remotes/origin/2.6.36-samsung

remotes/origin/HEAD -> origin/master

remotes/origin/master

這時會跑在no branch上，可以透過建立local的branch來存放2.6.24

$git branch 2.6.24

$git checkout 2.6.24

上面兩行合併就為

&git checkout -b 2.6.24

$git branch -a

這時你會看到2.6.24會有*號表示正在使用此branch，你可以多件幾個版本跟隨不同的local branch。

如果不小心建立的不要的branch可以透過下面指令來刪除

$git branch -d [deleted branch name]

大致上git就是這麼神奇，功能如此強大，就是剛開始clone的時間久了一點，因為他將從頭到尾的資料全部塞再.git底下

cross compile openssl

2011-02-27T23:35:00.000-08:00

download openssl from http://www.openssl.org/

The makefile is already exist, just modify Makefile without config. The modified configuration as shown below.

#you want to make install directory.

INSTALLTOP=/home/joe/test_program/openssl-1.0.0d/out

OPENSSLDIR=/home/joe/test_program/openssl-1.0.0d/out

CC= arm-none-linux-gnueabi-gcc

CFLAG= -O

DEPFLAG= -DOPENSSL_NO_GMP -DOPENSSL_NO_JPAKE -DOPENSSL_NO_MD2 -DOPENSSL_NO_RC5 -DOPENSSL_NO_RFC3779 -DOPENSSL_NO_STORE

PEX_LIBS=

EX_LIBS=

EXE_EXT=

ARFLAGS=

AR= arm-none-linux-gnueab-ar $(ARFLAGS) r

RANLIB= arm-none-linux-gnueab-ranlib

NM= nm

PERL= /usr/bin/perl

TAR= tar

TARFLAGS= --no-recursion

MAKEDEPPROG=makedepend

modified complete, type make, make install as usual.