Other Programming Topics
Timer
MiniGUI application can call function SetTimer
to create a timer. When the created timer is expired, the target window will receive message MSG_TIMER
, the identifier of the timer, and the system tick value that the timer is triggered. When the time is not needed, application program can call KillTimer
to delete the timer. SetTimerEx
and ResetTimerEx
are provided since MiniGUI version 2.0.4/1.6.10; they support timer callback procedure. The prototypes are as follow:
typedef BOOL (* TIMERPROC)(HWND, int, DWORD);
BOOL GUIAPI ResetTimerEx (HWND hWnd, int id, unsigned int speed,
TIMERPROC timer_proc);
BOOL GUIAPI SetTimerEx (HWND hWnd, int id, unsigned int speed,
TIMERPROC timer_proc);
#define SetTimer(hwnd, id, speed) \
SetTimerEx(hwnd, id, speed, NULL)
#define ResetTimer(hwnd, id, speed) \
ResetTimerEx(hwnd, id, speed, (TIMERPROC)0xFFFFFFFF)
The meaning of each parameter in TIMERPROC
is as follows:
The window handle passed when creating the timer. If no use, it can be any 32-bit value.
The timer ID.
The system tick value that the timer is triggered.
When TIMERPROC
return FALSE
, MiniGUI will delete the timer automatically. It can be used to create one-shot timer.
The timer mechanism of MiniGUI provides application with a comparatively convenient timing mechanism. However, the timer mechanism of MiniGUI has the following constraints:
Each message queue of MiniGUI-Threads only can manage 32 timers. Note that when creating a new thread, there will be a new message queue created to correspond, that is, each thread can have maximum 32 timers.
MiniGUI-Processes has only one message queue, which can manage maximum 32 timers.
The process of timer message is relatively special. Its implementation is similar to the signal mechanism of
UNIX
operating system. System will ignore this new timer message when a certain timer message has not been processed but new timer message occurs. It is because that if the frequency of certain timer is very high while the window processing this timer responds too slowly, when the message still needs to be posted, the message queue will be finally choked up.Timer message has the lowest priority. When there is not other type of messages in the message queue (such as posted message, notification message, and painting message), system will go to check if there is any expired timer.
When the timer frequency is very high, it is possible that such situation like timer message lost or interval asymmetry occurs. If application needs to use relatively accurate timer mechanism, it should use setitimer system call of Linux/UNIX with SIGALRM
signal. What needs to note is that the server process (mginit program) of MiniGUI-Processes has called setitimer system call to install the timer. Therefore, the program mginit
realized by application itself cannot use setitimer to realize the timer. However, the client program of MiniGUI-Processes still can call setitimer. MiniGUI-Threads has no such constraint.
Code in List 1 creates a timer with interval of one second, then use current time to set static text when timer is expired so as to display the clock. Finally, program will delete the timer while closing window.
List 1 Use of Timer
#define _ID_TIMER 100
#define _ID_TIME_STATIC 100
static char* mk_time (char* buff)
{
time_t t;
struct tm * tm;
time (&t);
tm = localtime (&t);
sprintf (buff, "%02d:%02d:%02d", tm->tm_hour, tm->tm_min, tm->tm_sec);
return buff;
}
static int TaskBarWinProc (HWND hWnd, int message, WPARAM wParam, LPARAM lParam)
{
char buff [20];
switch (message) {
case MSG_CREATE:
{
CreateWindow (CTRL_STATIC, mk_time (buff),
WS_CHILD | WS_BORDER | WS_VISIBLE | SS_CENTER,
_ID_TIME_STATIC, g_rcExcluded.right - _WIDTH_TIME - _MARGIN, _MARGIN,
_WIDTH_TIME, _HEIGHT_CTRL, hWnd, 0);
/* Create a timer with interval being one second.
* Its identifier is _ID_TIMER, target window is hWnd.
*/
SetTimer (hWnd, _ID_TIMER, 100);
break;
case MSG_TIMER:
{
/* Received a message of MSG_TIMER。
* Application should determine whether wParam is _ID_TIMER.
*/
SetDlgItemText (hWnd, _ID_TIME_STATIC, mk_time (buff));
break;
}
case MSG_CLOSE:
/* Delete the timer */
KillTimer (hWnd, _ID_TIMER);
DestroyAllControls (hWnd);
DestroyMainWindow (hWnd);
PostQuitMessage (hWnd);
return 0;
}
return DefaultMainWinProc (hWnd, message, wParam, lParam);
}
It is necessary to explain that the third argument of SetTimer
is used to specify the interval of timer. The default unit is 10 milliseconds.
Application can also call ResetTimer
to reset interval of a timer. The usage of the function is similar to SetTimer
.
Moreover, MiniGUI provides the other two functions to query system timer status. IsTimerInstalled
is used to check whether a timer is installed in the assigned window. HaveFreeTimer
is used to check whether system have free timer resources.
Clipboard
Clipboard is a tool to transfer data, and can be used for data communication among applications and application internals. Its principle is very simple, that is, a program places data on the clipboard, and another application takes the data from the clipboard. Clipboard is a data interchange station among applications.
Edit box control in MiniGUI supports clipboard operations, when the user selects text and press down CTRL+C keys, the data is copied to the text clipboard by default; when the users press down CTRL+V keys, the data is copied from the clipboard into edit box.
Creating and Destroying Clipboard
MiniGUI provides a default text clipboard, named as CBNAME_TEXT
(string name “text”), used for copying and pasting of text. Applications can use the clipboard directly without other additional operations. A clipboard defined by an application itself need to be created with CreateClipBoard
function, and to be destroyed with DestroyClipBoard
after using it.
There are at most NR_CLIPBOARDS
clipboards in MiniGUI, including the system default text clipboard and clipboards defined by user. NR_CLIPBOARDS
macro is defined to be 4 in minigui/window.h
header file by default.
CreateClipBoard
function creates a clipboard with specified name, this name can not be same as the name of existing clipboards (system defined or user defined):
int GUIAPI CreateClipBoard (const char* cb_name, size_t size);
Argument cb_name
specifies the name of a clipboard; argument size specifies the size of data to be stored by the clipboard. If creates successfully, the function returns CBERR_OK
; if the name is reduplicate, the function returns CBERR_BADNAME
; if the memory is not enough, the function returns CBERR_NOMEM
.
The DestroyClipBoard
function destroys a user defined clipboard created by CreateClipBoard
function:
int GUIAPI DestroyClipBoard (const char* cb_name);
Transferring Data to Clipboard
SetClipBoardData
function transfers data to a specified clipboard.
int GUIAPI SetClipBoardData (const char* cb_name, void* data, size_t n, int cbop);
Here, the argument cb_name
specifies the name of clipboard; data is the pointer to data buffer; n is the size of data; cbop is the operation type, which can be:
CBOP_NORMAL
: Default overwriting operation. The new data overwrites the existing data on the clipboard;CBOP_APPEND
: Appending operation. The new data will be appended to the existing data on the clipboard
Getting Data from Clipboard
GetClipBoardDataLen
function is used to get the size of data on clipboard.
size_t GUIAPI GetClipBoardDataLen (const char* cb_name);
GetClipBoardData
function is used to copy the data on clipboard to the specified data buffer:
size_t GUIAPI GetClipBoardData (const char* cb_name, void* data, size_t n);
Here the argument cb_name
specifies the name of the clipboard; data is the pointer to the data buffer; n specifies the size of specified data buffer. The function returns the size of gotten data from the clipboard
Generally speaking, you can use GetClipBoardDataLen
function to get the size of data before using GetClipBoardData
function to get data of clipboard, so that you can specify an appropriate data buffer to save data.
GetClipBoardByte
function is used to get a byte from the specified position of data on clipboard.
int GUIAPI GetClipBoardByte (const char* cb_name, int index, unsigned char* byte);
Here the argument index specifies the index position of specified data; byte is used to save the gotten byte data.
Reading/Writing Configuration File
The configuration file of MiniGUI (default as /usr/local/etc/MiniGUI.cfg file) uses Windows INI-like
file format. This format is very simple, seen as follows:
[section-name1]
key-name1=key-value1
key-name2=key-value2
[section-name2]
key-name3=key-value3
key-name4=key-value4
The information of such configuration file is grouped by section, then uses key=value format to appoint parameter and its value. Application can also use this format to store some configuration information. So, MiniGUI provides following functions (minigui/minigui.h):
int GUIAPI GetValueFromEtcFile (const char* pEtcFile, const char* pSection,
const char* pKey, char* pValue, int iLen);
int GUIAPI GetIntValueFromEtcFile (const char* pEtcFile, const char* pSection,
const char* pKey, int* value);
int GUIAPI SetValueToEtcFile (const char* pEtcFile, const char* pSection,
const char* pKey, char* pValue);
GHANDLE GUIAPI LoadEtcFile (const char* pEtcFile);
int GUIAPI UnloadEtcFile (GHANDLE hEtc);
int GUIAPI GetValueFromEtc (GHANDLE hEtc, const char* pSection,
const char* pKey, char* pValue, int iLen);
int GUIAPI GetIntValueFromEtc (GHANDLE hEtc, const char* pSection,
const char* pKey, int *value);
int GUIAPI SetValueToEtc (GHANDLE hEtc, const char* pSection,
const char* pKey, char* pValue);
int GUIAPI RemoveSectionInEtcFile (const char* pEtcFile, const char* pSection);
int GUIAPI GetValueFromEtcSec (GHANDLE hSect,
const char* pKey, char* pValue, int iLen);
int GUIAPI GetIntValueFromEtcSec (GHANDLE hSect,
const char* pKey, int* pValue);
int GUIAPI SetValueToEtcSec (GHANDLE hSect,
const char* pKey, char* pValue);
int GUIAPI SaveEtcToFile (GHANDLE hEtc, const char* file_name);
GHANDLE GUIAPI FindSectionInEtc (GHANDLE hEtc,
const char* pSection, BOOL bCreateNew);
int GUIAPI RemoveSectionInEtc (GHANDLE hEtc, const char* pSection);
The use of first three functions is as follows:
GetValueFromEtcFile
: To get a specified key value from a specified configuration file. The value of the key returns as a string.GetIntValueFromEtcFile
: To get a specified key integer value from a specified configuration file. This function converts the accepted string to integer value (using strtol function) and then returns the value.SetValueToEtcFile
: This function stores the given key value into a specified configuration file. If the configuration file does not exist, this function will create new configuration file. If the file exists, the old value will be covered.
The next five functions are new configuration file reading/writing function since MiniGUI version 1.6.x; the use of them is as follows:
LoadEtcFile
: Read a specified configuration file into memory and returns a configuration object handle, then the related functions can visit the configuration information in memory through this handle.UnloadEtcFile
: Release the configuration information in memory.GetValueFromEtc
: The way of use is similar toGetValueFromEtcFile
; But its first argument is the configuration object handle, not the file name. This function can be used to get configuration information from memory.GetIntValueFromEtc
: The way of use is similar toGetIntValueFromEtcFile
.SetValueToEtc
: Similar toSetValueToEtcFile
, but this function only changes the configuration key value in memory, does not affect the content in the file.
The last seven functions are new configuration file reading/writing function since MiniGUI version 2.0.4/1.6.10; the use of them is as follows:
RemoveSectionInEtcFile
: Remove a specified section from a specified configuration file.RemoveSectionInEtc
: Remove a specified section from the configuration information in memory.GetValueFromEtcSec
: Get value from a specified section in memory. Similar toGetValueFromEtc
.GetIntValueFromEtcSec
: Get an integer value from specified section in memory. Similar toGetIntValueFromEtc
.SetValueToEtcSec
: Set the value in a specified section in memory. Similiar toSetValueToEtc
andSetValueToEtc
.FindSectionInEtc
: Find or create a specified section from the configuration information in memory.SaveEtcToFile
: Save the configuration information in memory into a specified file.
These functions are usually used to read all information of configuration file for once. When need to get relatively more key values, this function will first use LoadEtcFile
to read in a configuration file, and then use GetValueFromEtc
to get key value. When there is no need to visit configuration information, use UnloadEtcFile
to release the configuration object handle.
Assuming that certain configuration file records some application information, and has the following formats:
[mginit]
nr=8
autostart=0
[app0]
path=../tools/
name=vcongui
layer=
tip=Virtual&console&on&MiniGUI
icon=res/konsole.gif
[app1]
path=../bomb/
name=bomb
layer=
tip=Game&of&Minesweaper
icon=res/kmines.gif
[app2]
path=../controlpanel/
name=controlpanel
layer=
tip=Control&Panel
icon=res/kcmx.gif
The section [mginit] records the number of applications and its automatically-startup index (autostart key). The section [appX] records information of each application program, including the path, the name, and the icon of the application. Code in List 2 illustrates how to use the functions above to get such information (the code comes from program mginit
of mg-samples).
List 2 Using MiniGUI configuration file functions to get information
#define APP_INFO_FILE “mginit.rc”
static BOOL get_app_info (void)
{
int i;
APPITEM* item;
/* Get information of the number of programs */
if (GetIntValueFromEtcFile (APP_INFO_FILE, "mginit", "nr", &app_info.nr_apps) = ETC_OK)
return FALSE;
if (app_info.nr_apps <= 0)
return FALSE;
/* Get index of autostarting application */
GetIntValueFromEtcFile (APP_INFO_FILE, "mginit", "autostart", &app_info.autostart);
if (app_info.autostart >= app_info.nr_apps || app_info.autostart < 0)
app_info.autostart = 0;
/* Calloc inforamtion structure of application */
if ((app_info.app_items = (APPITEM*)calloc (app_info.nr_apps, sizeof (APPITEM))) == NULL) {
return FALSE;
}
/* Get information of each application such as path, name and icon*/
item = app_info.app_items;
for (i = 0; i < app_info.nr_apps; i++, item++) {
char section [10];
sprintf (section, "app%d", i);
if (GetValueFromEtcFile (APP_INFO_FILE, section, "path",
item->path, PATH_MAX) = ETC_OK)
goto error;
if (GetValueFromEtcFile (APP_INFO_FILE, section, "name",
item->name, NAME_MAX) = ETC_OK)
goto error;
if (GetValueFromEtcFile (APP_INFO_FILE, section, "layer",
item->layer, LEN_LAYER_NAME) = ETC_OK)
goto error;
if (GetValueFromEtcFile (APP_INFO_FILE, section, "tip",
item->tip, TIP_MAX) = ETC_OK)
goto error;
strsubchr (item->tip, '&', ' ');
if (GetValueFromEtcFile (APP_INFO_FILE, section, "icon",
item->bmp_path, PATH_MAX + NAME_MAX) = ETC_OK)
goto error;
if (LoadBitmap (HDC_SCREEN, &item->bmp, item->bmp_path) = ERR_BMP_OK)
goto error;
item->cdpath = TRUE;
}
return TRUE;
error:
free_app_info ();
return FALSE;
}
If using LoadEtcFile
, GetValueFromEtc
, and UnloadEtcFile
to implement above example, the code will be as follows:
GHANDLE hAppInfo;
HAppInfo = LoadEtcFile (APP_INFO_FILE);
//…
get_app_info ();
//…
UnloadEtcFile (hAppInfo);
We also need change GetValueFromEtcFile
of function get_app_info
to GetValueFromEtc
.
Writing Portable Program
As we know, the CPU
used by most embedded system has totally different construction and characteristic from the CPU
of normal desktop PC
. However, operating system and advanced language can hide these differences to a great extent. With the support of advanced language programming, the compiler and operating system can help programs solve most problems related to CPU
architecture and characteristic in order to save developing time and increase developing efficiency. However, application programs have to face some certain CPU
characteristics; the following aspects need to be paid more attention:
The order of byte. Generally, when
CPU
stores integer data of multi-bytes, it will store the low-bit byte in low address unit, such as Intel x86 series. SomeCPU
uses opposite order to store. For example, the popularly usedPowerPC
in embedded system stores low-bit byte in high address unit. The former is called little-endian system while the latter is called big-endian system.The Linux kernel on some platforms may lack of some advanced system calls, the most popular one is the system calls related to virtual memory mechanism. The Linux system running on certain
CPU
cannot provide virtual memory mechanism because of the limitation ofCPU
capability. For example,CPU
lack ofMMU
unit cannot provide the sharing memory of System VIPC
mechanism.
In order to make the portable code have most popular adaptability, application programs must notice these differences and write code according to different situations. Here we will describe how to write portable code in MiniGUI applications.
Using Endian-Specific Read/Write Functions of MiniGUI
In order to solve the first problem mentioned above, MiniGUI provides several endian-related read/write functions. These functions can be divided into two categories:
Functions used to swap the order of byte, including
ArchSwapLE16
,ArchSwapBE16
and so on.Functions used to read/write standard I/O stream, including
MGUI_ReadLE16
,MGUI_ReadBE16
and so on.
The first category is used to convert the 16-bit, 32-bit, or 64-bit integer into system native byte from certain byte order. For example:
int fd, len_header;
...
if (read (fd, &len_header, sizeof (int)) == -1)
goto error;
#if MGUI_BYTEORDER == MGUI_BIG_ENDIAN
len_header = ArchSwap32 (len_header); // If it is big-endian system, swap the order
#endif
...
The above code first uses read system call to read an integer value from the a specified file descriptor to variable len_header.
The integer value saved in this file is in little-endian, so the byte order of this integer value has to be swapped if this integer value is used in big-endian system. We can use ArchSwapLE32
to convert 32-bit integer value into system native byte order. Also, we can swap the bytes only for big-endian system, and then we just need to use ArchSwap32
function.
The functions (or macro) used to swap bytes are as follow:
ArchSwapLE16(X
) converts the specified 16-bit integer value (stored in little endian byte order) to system native integer value. If system is little endian, this function will directly return X; if system is big endian, the function will callArchSwap16
to swap the bytes.ArchSwapLE32(X
) converts the specified 32-bit integer value (stored in little endian byte order) to system native integer value. If system is little endian, this function will directly return X; if system is big endian, the function will callArchSwap32
to swap the bytes.ArchSwapBE16(X
) converts the specified 16-bit integer value (stored in big endian byte order) to system native integer value. If system is big endian, this function will directly return X; if system is little endian, the function will callArchSwap16
to swap the bytes.ArchSwapBE32(X
) converts the specified 32-bit integer value (stored in big endian byte order) to system native integer value. If system is big endian, this function will directly return X; if system is little endian, the function will callArchSwap32
to swap the bytes.
The second category of functions provided by MiniGUI is used to read/write integer value from standard I/O file object. If the file is stored in little endian byte order, the function uses MGUI_ReadLE16
and MGUI_ReadLE32
to read integer value by converting integer value to system native byte order, whereas uses MGUI_ReadBE16
and MGUI_ReadBE32
. If the file is stored as little endian byte order, the function will use MGUI_WriteLE16
and MGUI_WriteLE32
to write integer value after converting integer value from system native byte order to little endian; whereas use MGUI_WriteBE16
and MGUI_WriteBE32
. The following code explains the above functions:
FILE* out;
int count;
...
MGUI_WriteLE32 (out, count); // Write count to the file in little endian
...
Using Condition Compilation to Write Portable Code
When regarding problems related to portability, we can easily use the way described above to perform function wrap in order to provide well portable code. However, sometime we cannot use such way to provide the portable code, so we can only use conditional compilation. Code in List 3 illustrates how to use conditional compilation to ensure the program running well (the code come from MiniGUI src/kernel/sharedres.c).
List 3 The usage of conditional compilation
/* If system does not support memory share, define _USE_MMAP */
#undef _USE_MMAP
/* #define _USE_MMAP 1 */
void *LoadSharedResource (void)
{
#ifndef _USE_MMAP
key_t shm_key;
void *memptr;
int shmid;
#endif
/* Load share resource*/
...
#ifndef _USE_MMAP /* Get object of share memory*/
if ((shm_key = get_shm_key ()) == -1) {
goto error;
}
shmid = shmget (shm_key, mgSizeRes, SHM_PARAM | IPC_CREAT | IPC_EXCL);
if (shmid == -1) {
goto error;
}
// Attach to the share memory.
memptr = shmat (shmid, 0, 0);
if (memptr == (char*)-1)
goto error;
else {
memcpy (memptr, mgSharedRes, mgSizeRes);
free (mgSharedRes);
}
if (shmctl (shmid, IPC_RMID, NULL) < 0)
goto error;
#endif
/* Open a file */
if ((lockfd = open (LOCKFILE, O_WRONLY | O_CREAT | O_TRUNC, 0644)) == -1)
goto error;
#ifdef _USE_MMAP
/* If use mmap, write share resource into the file*/
if (write (lockfd, mgSharedRes, mgSizeRes) < mgSizeRes)
goto error;
else
{
free(mgSharedRes);
mgSharedRes = mmap( 0, mgSizeRes, PROT_READ|PROT_WRITE, MAP_SHARED, lockfd, 0);
}
#else
/* otherwise write the object ID of share memory into the file*/
if (write (lockfd, &shmid, sizeof (shmid)) < sizeof (shmid))
goto error;
#endif
close (lockfd);
#ifndef _USE_MMAP
mgSharedRes = memptr;
SHAREDRES_SHMID = shmid;
#endif
SHAREDRES_SEMID = semid;
return mgSharedRes;
error:
perror ("LoadSharedResource");
return NULL;
}
The MiniGUI-Processes server program to load sharing resource uses the above code fragment. If system supports shared memory, it will initialize the shared memory object and associate the shared resource with the shared memory object, then write the shared memory object ID into a file; if system does not support shared memory, it will write all initialized sharing resource into a file. If the system support shared memory, clients can get shared memory object ID from the file and directly attach it; if the system does not support shared memory, clients can use mmap system call to map the file to the address space of them. Code of clients can be seen in List 4.
List 4 The usage of conditional compilation (cont.)
void* AttachSharedResource (void)
{
#ifndef _USE_MMAP
int shmid;
#endif
int lockfd;
void* memptr;
if ((lockfd = open (LOCKFILE, O_RDONLY)) == -1)
goto error;
#ifdef _USE_MMAP
/* Use mmap to image share resource to process address space */
mgSizeRes = lseek (lockfd, 0, SEEK_END );
memptr = mmap( 0, mgSizeRes, PROT_READ, MAP_SHARED, lockfd, 0);
#else
/* Otherwise get ID of the object of share memroy, and associate the share memory */
if (read (lockfd, &shmid, sizeof (shmid)) < sizeof (shmid))
goto error;
close (lockfd);
memptr = shmat (shmid, 0, SHM_RDONLY);
#endif
if (memptr == (char*)-1)
goto error;
return memptr;
error:
perror ("AttachSharedResource");
return NULL;
}
Fixed-Point Computing
Usually when we perform math operations, we will use float-point to represent real number, and use <math.h> head file to calculate the float-point number. However, float-point calculation is a time-consuming process. Therefore, in order to reduce extra CPU
instructions caused by float-point calculation, some three-dimension graphics application always use fixed-point number to represent real number, which will greatly accelerate the calculation of three-dimension graphical rendering. MiniGUI also provides some fixed-point computing functions, divided into the following categories:
Conversion among integer, float-point number and fixed-point number. itofix converts integer to fixed-point number, while fixtoi converts fixed-point to integer; ftofix converts float-point number to fixed-point number, while fixtof converts fixed-point number to float-point number.
The basic arithmetic computing such as add, subtract, multiple, and divide of fixed-point numbers: fixadd, fixsub, fixmul, fixdiv, fixsqrt.
The triangle compute of fixed-point number: fixcos, fixsin, fixtan, fixacos, fixasin.
Matrix and vector computing. Matrix and vector related computing are important for three-dimension graphics. Readers can refer to
minigui/fixedmath.h
for the functions.
Code in List 5 illustrates the use of fixed-point number. This code converts plane rectangular coordinates to screen coordinates.
List 5 fixed-point computing
void scale_to_window (const double * in_x, const double * in_y, double * out_x, double * out_y)
{
fixed f_x0 = ftofix (get_x0());
fixed f_y0 = ftofix (get_y0());
fixed f_in_x = ftofix (*in_x);
fixed f_in_y = ftofix (*in_y);
fixed f_p = ftofix (get_pixel_length());
*out_x = fixtof(fixmul(fixsub(f_in_x, f_x0), f_p));
*out_y = -fixtof(fixmul(fixsub(f_in_y, f_y0), f_p));
}
The calculation of above program is very simple. The steps are as follow:
Converts the input parameters to fixed-point values.
Does the calculation by using the fixed-point values.
Converts the result values to float-point values.
<< Look and Fee and UI Effects | Table of Contents | Using mGUtils >>
Last updated