X-Git-Url: http://git.hungrycats.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=utils%2Fxshm.c;fp=utils%2Fxshm.c;h=c9498e5b0e18e39891744e845c27a6738db8afaa;hb=df7adbee81405e2849728a24b498ad2117784b1f;hp=0000000000000000000000000000000000000000;hpb=41fae2ad67bc37e31c4d967bae81e4f3f50fa55a;p=xscreensaver diff --git a/utils/xshm.c b/utils/xshm.c new file mode 100644 index 00000000..c9498e5b --- /dev/null +++ b/utils/xshm.c @@ -0,0 +1,192 @@ +/* xscreensaver, Copyright (c) 1993, 1994, 1995, 1996, 1997, 1998 + * by Jamie Zawinski + * + * Permission to use, copy, modify, distribute, and sell this software and its + * documentation for any purpose is hereby granted without fee, provided that + * the above copyright notice appear in all copies and that both that + * copyright notice and this permission notice appear in supporting + * documentation. No representations are made about the suitability of this + * software for any purpose. It is provided "as is" without express or + * implied warranty. + */ + +/* The MIT-SHM (Shared Memory) extension is pretty tricky to use. + This file contains the common boiler-plate for creating a shared + XImage structure, and for making sure that the shared memory segments + get allocated and shut down cleanly. + + This code currently deals only with shared XImages, not with shared Pixmaps. + It also doesn't use "completion events", but so far that doesn't seem to + be a problem (and I'm not entirely clear on when they would actually be + needed, anyway.) + + If you don't have man pages for this extension, see + http://www.physik.uni-regensburg.de/~scs22156/sofie-0.2/mit-shm.html + or in the R6 sources as "xc/doc/specs/Xext/mit-shm.ms", for example, + ftp://ftp.x.org/pub/R6.4/xc/doc/specs/Xext/mit-shm.ms + + (This document seems not to ever remain available on the web in one place + for very long; you can search for it by the title, "MIT-SHM -- The MIT + Shared Memory Extension".) + + To monitor the system's shared memory segments, run "ipcs -m". + */ + +#include "utils.h" + +#ifdef HAVE_XSHM_EXTENSION /* whole file */ + +/* #define DEBUG */ + +#include /* for perror() */ +#include /* for XDestroyImage() */ + +#include "xshm.h" +#include "resources.h" /* for get_string_resource() */ + +#ifdef DEBUG +# include +#endif + +extern char *progname; + + +/* The documentation for the XSHM extension implies that if the server + supports XSHM but is not the local machine, the XShm calls will return + False; but this turns out not to be the case. Instead, the server + throws a BadAccess error. So, we need to catch X errors around all + of our XSHM calls, sigh. + */ + +static Bool shm_got_x_error = False; +XErrorHandler old_handler = 0; +static int +shm_ehandler (Display *dpy, XErrorEvent *error) +{ + shm_got_x_error = True; + +#ifdef DEBUG + fprintf (stderr, "\n%s: ignoring X error from XSHM:\n", progname); + XmuPrintDefaultErrorMessage (dpy, error, stderr); + fprintf (stderr, "\n"); +#endif + + return 0; +} + + +#define CATCH_X_ERROR(DPY) do { \ + XSync((DPY), False); \ + shm_got_x_error = False; \ + if (old_handler != shm_ehandler) \ + old_handler = XSetErrorHandler (shm_ehandler); \ +} while(0) + +#define UNCATCH_X_ERROR(DPY) do { \ + XSync((DPY), False); \ + if (old_handler) \ + XSetErrorHandler (old_handler); \ + old_handler = 0; \ +} while(0) + + +XImage * +create_xshm_image (Display *dpy, Visual *visual, + unsigned int depth, + int format, char *data, + XShmSegmentInfo *shm_info, + unsigned int width, unsigned int height) +{ + Status status; + XImage *image = 0; + if (!get_boolean_resource("useSHM", "Boolean")) + return 0; + + if (!XShmQueryExtension (dpy)) + return 0; + + CATCH_X_ERROR(dpy); + image = XShmCreateImage(dpy, visual, depth, + format, data, shm_info, width, height); + UNCATCH_X_ERROR(dpy); + if (shm_got_x_error) + return 0; + +#ifdef DEBUG + fprintf(stderr, "\n%s: XShmCreateImage(... %d, %d)\n", progname, + width, height); +#endif + + shm_info->shmid = shmget(IPC_PRIVATE, + image->bytes_per_line * image->height, + IPC_CREAT | 0777); +#ifdef DEBUG + fprintf(stderr, "%s: shmget(IPC_PRIVATE, %d, IPC_CREAT | 0777) ==> %d\n", + progname, image->bytes_per_line * image->height, shm_info->shmid); +#endif + + if (shm_info->shmid == -1) + { + char buf[1024]; + sprintf (buf, "%s: shmget failed", progname); + perror(buf); + XDestroyImage (image); + image = 0; + XSync(dpy, False); + } + else + { + shm_info->readOnly = False; + image->data = shm_info->shmaddr = shmat(shm_info->shmid, 0, 0); + +#ifdef DEBUG + fprintf(stderr, "%s: shmat(%d, 0, 0) ==> %d\n", progname, + shm_info->shmid, (int) image->data); +#endif + + CATCH_X_ERROR(dpy); + status = XShmAttach(dpy, shm_info); + UNCATCH_X_ERROR(dpy); + if (shm_got_x_error) + status = False; + + if (!status) + { + fprintf (stderr, "%s: XShmAttach failed!\n", progname); + XDestroyImage (image); + XSync(dpy, False); + shmdt (shm_info->shmaddr); + image = 0; + } +#ifdef DEBUG + else + fprintf(stderr, "%s: XShmAttach(dpy, shm_info) ==> True\n", progname); +#endif + + XSync(dpy, False); + + /* Delete the shared segment right now; the segment won't actually + go away until both the client and server have deleted it. The + server will delete it as soon as the client disconnects, so we + should delete our side early in case of abnormal termination. + (And note that, in the context of xscreensaver, abnormal + termination is the rule rather than the exception, so this would + leak like a sieve if we didn't do this...) + + #### Are we leaking anyway? Perhaps because of the window of + opportunity between here and the XShmAttach call above, during + which we might be killed? Do we need to establish a signal + handler for this case? + */ + shmctl (shm_info->shmid, IPC_RMID, 0); + +#ifdef DEBUG + fprintf(stderr, "%s: shmctl(%d, IPC_RMID, 0)\n\n", progname, + shm_info->shmid); +#endif + } + + return image; +} + +#endif /* HAVE_XSHM_EXTENSION */