Description:
------------
I'm using the ondrej-ppa packages. This bug is triggered in a production setup with many requests and it does not seem to be triggered by specific requests. This is triggered in php-fpm and does not seem to be connected to a specific request.
We get the segfaults every few hours. I can imagine this bug gets triggered because RSHUTDOWN is not run, since this would set the DATEG(tzcache) = NULL; and thus not trigger the if condition in zm_activate_date:
	if (DATEG(timezone)) {
		efree(DATEG(timezone));
	}

Extensions loaded:
mysqlnd.so
opcache.so
pdo.so
xml.so
apcu.so
calendar.so
ctype.so
curl.so
dom.so
exif.so
fileinfo.so
ftp.so
gd.so
geoip.so
gettext.so
iconv.so
igbinary.so
intl.so
json.so
mbstring.so
mysqli.so
pdo_mysql.so
phar.so
posix.so
readline.so
redis.so
shmop.so
simplexml.so
sockets.so
sysvmsg.so
sysvsem.so
sysvshm.so
tideways.so
tokenizer.so
wddx.so
xmlreader.so
xmlwriter.so
xsl.so
zip.so
apc.so


Expected result:
----------------
No segfault

Actual result:
--------------
GDB Backtrace:
#0  zend_mm_free_heap (ptr=0x7fcab4135270, heap=0x7fcacd400040) at ./Zend/zend_alloc.c:1388
#1  _efree (ptr=0x7fcab4135270) at ./Zend/zend_alloc.c:2456
#2  0x000055b8b5a60795 in zm_activate_date (type=<optimized out>, module_number=<optimized out>) at ./ext/date/php_date.c:729
#3  0x000055b8b5c1b7b0 in zend_activate_modules () at ./Zend/zend_API.c:2600
#4  0x000055b8b5bae7e6 in php_request_startup () at ./main/main.c:1694
#5  0x000055b8b5a5f870 in main (argc=<optimized out>, argv=<optimized out>) at ./sapi/fpm/fpm/fpm_main.c:1927