aniani/elinks
1
0
Fork 0
mirror of https://github.com/rkd77/elinks.git synced 2024-09-27 02:56:18 -04:00
Code

bug 638: More comments. Assert that calls don't nest.

Browse Source
This commit is contained in:
Kalle Olavi Niemitalo 2008-03-22 13:58:36 +02:00 committed by Kalle Olavi Niemitalo
parent 4eefa983de
commit aae5788cc7
2 changed files with 50 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
doc/Doxyfile.in
View File

@ -228,6 +228,7 @@ PREDEFINED = "LIST_OF(element_T)=element_T list" \
CONFIG_ECMASCRIPT_SEE \
CONFIG_DEBUG \
CONFIG_UTF8 \
HAVE_SETENV \
HAVE_VARIADIC_MACROS
EXPAND_AS_DEFINED = LIST_HEAD \
INIT_LIST_OF

54
src/mime/backend/mailcap.c
View File

@ -647,15 +647,56 @@ get_mailcap_entry(unsigned char *type)
}
#if defined(HAVE_SETENV) || defined(HAVE_PUTENV)
/* restore == 0 set DISPLAY
* restore == 1 restore DISPLAY
*/
/** Set or unset the DISPLAY environment variable before a mailcap
* check, or restore it afterwards.
*
* In a mailcap file, each entry can specify a test command that
* checks whether the entry is applicable to the user's environment.
* For example:
*
* @verbatim
* audio/mpegurl; xmms %s; test=test "$DISPLAY" != ""
* @endverbatim
*
* This means the entry should be used only if the DISPLAY environment
* variable is not empty, i.e. there is an X display. In ELinks,
* check_entries() runs these test commands, so they inherit the
* environment variables of the master ELinks process. However, if
* the user is running ELinks on multiple terminals, then each slave
* ELinks process has its own environment variables, which may or may
* not include DISPLAY. Because the actual mailcap command may be run
* from a slave ELinks process and inherit the environment from it,
* any test command should also be run in the same environment.
*
* This function does not fully implement the ideal described above.
* Instead, it only sets the DISPLAY environment variable as ":0" if
* the terminal has any X display at all, or unsets DISPLAY if not.
* This should be enough for most test commands seen in practice.
* After the test commands of mailcap entries have been run, this
* function must be called again to restore DISPLAY.
*
* @todo Retrieve all environment variables from the slave process and
* propagate them to the test commands. Actually, it might be best
* to fork the test commands from the slave process, so that they
* would also inherit the controlling tty. However, that would
* require changing the interlink protocol and might risk deadlocks
* or memory leaks if a slave terminates without responding.
*
* @param xwin
* Whether the terminal has an associated X display.
* @param restore
* If this is 0, the function sets or clears DISPLAY, as described above.
* If this is 1, the function restores the original value of DISPLAY.
* There is only room for one saved value; do not nest calls. */
static void
set_display(int xwin, int restore)
{
static unsigned char *display = NULL;
if (!restore) {
assert(display == NULL);
if_assert_failed mem_free(display);
display = getenv("DISPLAY");
if (display) display = stracpy(display);
if (xwin) {
@ -677,13 +718,16 @@ set_display(int xwin, int restore)
setenv("DISPLAY", display, 1);
#else
{
static unsigned char DISPLAY[1024] = { 'D','I','S','P','L','A','Y','=' };
static unsigned char DISPLAY[1024] = "DISPLAY=";
/* DISPLAY[1023] == '\0' and we don't let
* strncpy write that far, so putenv always
* gets a null-terminated string. */
strncpy(DISPLAY + 8, display, 1023 - 8);
putenv(DISPLAY);
}
#endif
mem_free(display);
mem_free_set(&display, NULL);
} else {
#ifdef HAVE_UNSETENV
unsetenv("DISPLAY");