Manual for snprintf says:
If the output was truncated, the return value is -1, oth-
erwise it is the number of characters stored, not includ-
ing the terminating null.
And that's a blatant lie - in reality, libc 2.1 always returns number of
characters that _would be_ stored. I hate those libc bugs. Now we should go
and check every place we trusted snprintf return value.
(the older code worked, even though it didn't compile well for some very odd reason, probably
a compiler bug; The new code would cause PHP to crash if both display_errors/log_errors and
track_errors were enabled)
Added PHP API for registering tick functions. Ticks is a per-instruction
callback in Zend that allows us to implement cooperative multitasking
within PHP. Example:
In extension code:
php_add_tick_function(my_tick_func);
void my_tick_func(int count) {
php_printf("[tick %d]", count);
}
In PHP code:
<?php
declare(ticks = 1) {
print "there should be two ticks here.\n";
}
?>
- but as it is Sascha who has reverted a zillion patches I don't feel that
- bad about it :)
- The patch is morally incorrect and it actually also has a crash bug which
- I won't point out because it shouldn't be there.
- SG(path_translated) should be set by the SAPI module correctly and if it
- isn't the SAPI module should be fixed and not add code to PHP. There is
- enough special case code in PHP and I would like to clean it up and not
- add more.
- Last but not least, old_cwd takes 4KB of stack space. We should probably
- malloc() it because in ISAPI we only have 10KB of stack space.
to the directory where the executing script is located.
Since this needs to be implemented for all SAPI modules anyway, this
change moves the functionality to php_execute_script() and gets rid
of the per-module code.
long-requested functionality, now that output buffering is re-entrant:
function eval_ret($code)
{
ob_start();
eval($code);
$retval = ob_get_contents();
ob_end_clean();
return $retval;
}
plus a little cleanup and rearranging in command line option parsing
@ CGI aka. command line version has now an option '-l' for syntax check
@ without execution (Hartmut)
- So here is the short version:
- a) Start moving to binary opens in Windows
- b) Give checkuid_mode() a small face lift including the fopen-wrappers.c
- The mode to this function should at least be a #define but that is for
- another day. Anyway this whole stuff should be given more face lifts in
- the future.
