问题
NCurses appears to be a popular library. One of its weaknesses is, that it is not threadsafe. It should not be hard to wrap shared ressources in mutexes.
Is there a specific reason, why noone has started a threadsafe branch? (Legal issues, introducing a platform dependency, ...)
Edit: I do not mean the use_screen or use_window functions. These apparently require the user to change his NCurses-based code. It should be possible to add a mutex to the shared resources within the NCurses itself, and all accessing functions acquire the mutex before doing something with the window. I am imagining something like this within NCurses:
#if __cplusplus >= 201103L
#include <mutex>
#define THREADSAFE
#endif
...
#ifdef THREADSAFE
std::recursive_mutex mxCurscr;
#endif
...
int doupdate(void)
{
#ifdef THREADSAFE
mxCurscr.lock();
#endif
... // <-- Access the screen here.
#ifdef THREADSAFE
mxCurscr.unlock()
#endif
}
- This does not rely on anything but the C++11 standard.
- This is compatible with older compilers. (But no threadsafety then.)
- It should not take more than one or two days to make the modifications.
- This satisfies the demand for a threadsafe NCurses.
- The user of the NCurses library will not have to bother.
- The work is being done one time for all users, instead of having every user implementing its own thread-safety.
So, where is the catch?
回答1:
It's already been done (in ncurses 5.7, released November 2008), but not much used. See the curs_threads manual page for instance. It is not a feature in the default configuration because it
- changes the ABI (application binary interface), and
- adds some restrictions on how standard variables are used.
回答2:
Thread-safe code looks like this.
#include <stdio.h>
#include <ncurses.h>
#include <pthread.h>
#include <time.h>
#include <unistd.h>
pthread_mutex_t MUTEX; /* mutex for sync display */
#define LOCK pthread_mutex_lock(&MUTEX)
#define UNLOCK pthread_mutex_unlock(&MUTEX)
typedef struct _WIN_struct {
int startx, starty;
int height, width;
} WIN;
WIN winTitle; /* title win */
WIN winMenu; /* Main menu win */
WIN winNews; /* win news */
WINDOW *create_newwin(int height, int width, int starty, int startx) {
WINDOW *local_win;
local_win = newwin(height, width, starty, startx);
box(local_win, 0, 0);
wrefresh(local_win);
return local_win;
}
char *getTimeNow() {
time_t rawtime;
struct tm *timeinfo;
time(&rawtime);
timeinfo = localtime(&rawtime);
return asctime(timeinfo);
}
void *threadfunc_title(void *p) {
WINDOW *windowTitle;
LOCK;
windowTitle = create_newwin(winTitle.height, winTitle.width, winTitle.starty, winTitle.startx);
UNLOCK;
/* show title and time */
for (;;) {
sleep(1);
}
}
void *threadfunc_menu(void *p) {
WINDOW *windowMenu;
LOCK;
windowMenu = create_newwin(winMenu.height, winMenu.width, winMenu.starty, winMenu.startx);
UNLOCK;
for (;;) {
/* now do nothing */
sleep(1);
}
}
void *threadfunc_news(void *p) {
WINDOW *windowNews;
LOCK;
windowNews = create_newwin(winNews.height, winNews.width, winNews.starty, winNews.startx);
UNLOCK;
for (;;) {
sleep(1);
}
}
void initWin(WIN *p_win, int height, int width, int starty, int startx) {
p_win->height = height;
p_win->width = width;
p_win->starty = starty;
p_win->startx = startx;
}
int main(int argc, char *argv[])
{
pthread_t pidTitle;
pthread_t pidMenu;
pthread_t pidNews;
initscr();
start_color();
cbreak();
keypad(stdscr, TRUE);
noecho();
/* init location */
initWin(&winTitle, LINES*0.25, COLS, 0 , 0);
initWin(&winMenu, LINES*0.75, COLS*0.60, LINES*0.25, 0);
initWin(&winNews, LINES*0.75, COLS*0.40, LINES*0.25, COLS*0.60);
pthread_create(&pidTitle, NULL, threadfunc_title, NULL);
pthread_create(&pidMenu, NULL, threadfunc_menu, NULL);
pthread_create(&pidNews, NULL, threadfunc_news, NULL);
pthread_join(pidTitle, NULL);
pthread_join(pidMenu, NULL);
pthread_join(pidNews, NULL);
endwin();
return 0;
}
来源:https://stackoverflow.com/questions/29910562/why-has-no-one-written-a-threadsafe-branch-of-the-ncurses-library