Ting 이네 Software 이야기

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

Let's consider two different level - Coder, Programmer.
Coder is engineer who can modify and implement not-big-size source code under the given SW architecture, requirement and environment.
Programmer is engineer who can define requirements, design SW architecture, analysis and fix system-wide-issues.

At first, we decide which level of SW engineer should be recruited.

To recruit Coder, following questions may be enough.
- Questions about language syntax and it's application.
- Simple problem-solving-ability (ex. simple optimization and so on.)
- Coding test. (ex. implement linked list, Queue or Stack etc)

But, to recruit Programmer, you may need additional questions to know about his/her programming policy and values.
For example,
- What's your opinion about 'modularity'?
- What is 'software layer' and what it should be?
- Please tell me about 'information hiding'

But, someone may give well-known answer that may come from books. So, detail and concrete question should be follows.
For example, followings can be asked after question "What's your opinion about 'modularity'"
- Then, What should good modular architecture be?
- Show example of good modular architecture.
- What kind of design can be used to connect between modules.
- What is good interface design to support modularity?
- Please design linked list with C language focusing on modularity.

... I am not sure that this kind of evaluation can be possible in practical. But...

VC[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

__stdcall vs. __cdecl

refer below page..

Sometimes, so-called 'context' data structure is used to handling set of related information. And this context data tends to be required as parameter by various functions. Here is pros and cons.

Pros:
- It is easy to handle lots of related data - data allocation and free etc.
- We can save costs of parameter passing; Passing context is enough.

Cons:
- It is difficult to know which one is input and which one is output, because whole context is passed as parameter. This decrease readability very much.
- Usually, context is passed as parameter to lots of functions. So, just like 'global' data, it is very difficult to trace the value-change-history. That is, debugging/maintaining becomes difficult.

One good way to using context structure; Only very few modules can change/modify context data. And others only can read this one. Than, lot's of issues in "Cons" can be overcome.

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

* Making Blended & Clink-Through window

We can use following window style option at CreateWindowEx().
WS_EX_TRANSPARENT -> click-though
WS_EX_LAYERED -> alpha-blended window.

We can changne configuration in runtime by using following API.

SetLayeredWindowAttributes(...)
: Layered window's attributes (Alpha value, Transparet color etc) can be changed.

SetWindowLong(...)
: Window의 Ex Style can be changed. We can set or clear "Click-Through" by setting or clearing WS_EX_TRANSPARENT attributes.

* We can remove menu

by set 'lpszMenuName' as 'NULL' at 'WNDCLASSEX' structure when 'CreateWindowEx' is called.

* We can change initial background color of Window

by creating brush of preferred color at 'hbrBackground'.

* We can remove 'titlebar' and 'boarder'

by set 'dwStype' parameter as 'WS_POPUP' style.

* (Issue) Maximized window covers Taskbar!

We can resolve this issue by using 'WM_GETMINMAXINFO' and 'SystemParametersInfo/SPI_GETWORKAREA'.
('WM_GETMINMAXINFO' is sent when Window is moved or Window's size is changed.)

Let's put following code in the handler of 'WM_GETMINMAXINFO'.

MINMAXINFO* pi = (MINMAXINFO*)lParam;
RECT        r;
SystemParametersInfo(SPI_GETWORKAREA, 0, &r,0);
memset(pi, 0x00, sizeof(MINMAXINFO));
pi->ptMaxSize.x = r.right - r.left;
pi->ptMaxSize.y = r.bottom - r.top;
pi->ptMaxPosition.x = r.left;
pi->ptMaxPosition.y = r.top;

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

Win32에서 IME가 Active되어 있으면, 즉 Imm context가 window handle에 associate되어 있고, 한글 입력모드에서 입력을 하고 있으면, WM_KEYDOWN의 wParam이 VK code가 올라 오지 않고, VK_PROCESSKEY가 올라온다. 즉 Imm이 해당 key를 processing하고 있다는 뜻이다. 여기서 어떻게 원래의 Key code를 얻어 올 수 있을까?

MSDN을 참조하면, "mmGetVirtualKey()"를 쓰면 된다고 하는데, 막상 써 보면, VK_PROCESSKEY가 그대로 return된다. 웹을 더 찾아보면, TranslateMessage를 호출하기 전에 사용하면 된다고 하는데, 안해 봐서 모르겠고, 또 그렇게 하기도 힘들다.

이럴때는 약간은 portability와 compatibility가 희생해서, scan code를 사용하는 방법이 있다.
몇몇 platform의 경우, (Embedded환경은 특히...) scan code가 전부 0으로 올라오는 경우도 있어서 100% portability가 떨어지긴 하나 어차피 100% portable은 존재하지 않으니까... :-)

scan code는 WM_KEYDOWN message의 lParam의 16~23 bit (8bit)인데, 이를 "MapVirtualKey()"함수를 사용해서 virtual key code를 얻을 수 있다.
뭐 일단 얼마나 portability가 희생되는지는 모르겠으나, 어느정도 잘 동작하는 것은 확인했다. 이 방법을 사용하는 것도 괜찮겠다.

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

"WM_IME_COMPOSITION" 을 사용해야 한다.
뭐 MSDN을 찾아보면 관련 함수도 많이 나오고 설명도 많이 나오니 패스.
Font를 찍을 때는 TrueType의 경우, 그냥 GetWidth32W (32Bit 환경)을 써서 font width를 계산하면 된다.

몇 가지 주의할 점만 짚어 보겠다.
먼저 조합도중 Space입력.
이 넘이 상당히 많은 것을 자동으로 해 주어서 편하긴 한데, Asian국가중 힘 없는 한글은 상당히 무시되어 있다. ㅜ.ㅜ
젤 큰 문제는, 한글 조합도중 'Space'를 누르면, space가 Keycode로 올라와야 하는데, 이 Space는 그냥 "VK_PROCESSKEY"로 올라오고, "WM_IMM_COMPOSITION"에는 result로 space가 입력되기 전까지 조합되었던 글자가 올라온다. 즉, "WM_IME_COMPOSITION"만을 이용해서 한글 입력기를 만들게 되면, '가'를 입력하고(조합중) space를 치면, '가 '가 아니라 '가' 가 찍힌다. (Space가 나타나지 않는다.)
이를 처리해야 한다.!! (사실 일본어나 중국어의 경우 space는 character table을 보여주는 key로 사용되는 듯 하다. 그렇지만, 한글에는 그게 필요없으니.. MS한테 무시당했다...-_-;)

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

Have you ever wanted to compile Linux binaries under Windows? This is how it

happens:
1. Get Cygwin setup.exe: http://www.cygwin.com/
1.1. Run setup.exe and continue to package selection list.
1.2. Under Devel catagory select tools you need for compiling your source. For example 'GNU make'.
1.3. Finish installing.

2. Get linux cross-compilers for cygwin:
"cygwin-gcc-linux.tar.bz2" (68.2 Mb). md5sum: 340e91a346f5bb17e660db10e43005b8
These compilers are made with crosstool 0.28-rc37. This package contains:
gcc-3.3.4 and gcc-2.95.3 for i386 (glibc 2.1.3) and gcc-3.3.3 for amd64
(glibc 2.3.2).

--------------------------------------------------------------------------------

Note! There is now newer version of GCC avaible with glibc 2.3.2:
"cygwin-gcc-3.3.6-glibc-2.3.2-linux.tar.bz2 (i386, x86_64)".
2.1. Copy 'cygwin-gcc-linux.tar.bz2' to 'c:\cygwin' or install directory which you selected in setup.exe.
2.2. Open Cygwin shell and change directory to root with 'cd /'.
2.3. Uncompress to Cygwin root with command:
'tar -jxvf cygwin-gcc-linux.tar.bz2'.

Cross-compilers are installed under '/opt/crosstool'. You can use them directly or with commands: gcc-linux, g++-linux, gcc-linux-2.95, g++-linux-2.95, gcc-linux-x86_64 and g++-linux-x86_64.

source : http://metamod-p.sourceforge.net/cross-compiling.on.windows.for.linux.html

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

Error when executing shell command in "Cygwin make"
Here is the case.

path=c:\cygwin\bin; (...omitted... );; (...omitted...)

Take you attention to ';;' - two consecutive semicolon.

In above path environment, executable binaries like 'cp', 'mkdir' that are in 'c:\cywin\bin', works well. But, shell command like 'cmd' raise error something like "command not found".

After removing ';;' - consecutive semicolon (empty path) -, all works well.

So, here is my inference.

"Cywin make" executes command by following order.
  1. searching path environment to file executable binary.
  2. regards command as 'shell command' and try to execute.

In case that ';;' is appeared in the path, "Cygwin make" cannot recognize path after ';;'. So, executables located in the directory those are before ';;' works well. But, those are not, "Cygwin make" treats it as "shell command" and raises error (there is no shell command like this...).

[[ blog 이사 과정에서 정확한 posting날짜가 분실됨. 년도와 분기 정도는 맞지 않을까? ]]

RVCT and ADS make long jump be possible by adding veneer. But, interestingly, link tool of RVCT and ADS, added veneer only when jump to the code located in other execution area. So, in the case that long jump to the same-execution-area-code is needed, linker generates "memory relocation error" in ADS and RVCT. That's why we need to classify code and put them into several execution area.

TI's TMS470 adds veneer('Trampoline' in TMS) even if target code of long jump is in same execution area. But TMS470 is very slow... :-(

I have no idea about gnu arm eabi tools...