Marcus Boerger [Tue, 12 Nov 2002 20:56:47 +0000 (20:56 +0000)]
Implemented -n switch to skip parsing ini at startup as suggested by Wez.
#The switch 'n' was planned to be used for beautifying....delete n to make
#clear these functions do not have a switch yet.
HTTP_RAW_POST_DATA BC fixes
# hopefully all done, commiting anyway to continue work on my home box
php://input stream fixes (POST data handerl mangles data, CLI crashbug)
Ilia Alshanetsky [Tue, 12 Nov 2002 14:40:00 +0000 (14:40 +0000)]
Changed max_input_time PHP_INI_SYSTEM|PHP_INI_PERDIR because ini_set() will
already be too late, the POST/GET/COOKIE processing occures before the script
gets parsed.
Changed the default value to -1. If users have previously (older PHPs)
adressed the problem by setting timeout to a large value, we can avoid breaking
their script by detecting the -1 and using timeout_seconds instead of
max_input_time when setting the 'input' timeout.
Timm Friebe [Tue, 12 Nov 2002 13:13:12 +0000 (13:13 +0000)]
- Added optional parameter to sybase_fetch_object() (Timm)
- Changed php_error and zend_error to php_error_docref (Timm)
- Changed sybase_query() to clean up and free unread results (Timm)
Marcus Boerger [Tue, 12 Nov 2002 11:49:11 +0000 (11:49 +0000)]
new function gd_info() returns an associative array of gd support options. # The index names are those from the minfo function. Therefore it is
# easy to look into phpinfo() to see which index names are possible.
Yasuo Ohgaki [Tue, 12 Nov 2002 09:07:36 +0000 (09:07 +0000)]
Speed up pg_escape_bytea, from O(n^2) => O(n). My test script speed up
from 50 sec to 5 sec. (sscanf is too slow at least under my linux, I
might write better function that unescape values later)
Use emalloc/erealloc for bundled PQescapeBytea(php_pgsql_escape_bytea)
Zak Greant [Tue, 12 Nov 2002 01:41:16 +0000 (01:41 +0000)]
Modified php_mysql_do_connect() as suggested by Nick Gaugler - using
mysql_ping() as a more efficient alternative to using mysql_stat() to
check if the server is alive and then calling mysql_(real_)?connect() to
reconnect.
Simple tests of opening pconnects indicate that only about 10k of data per
ping needs to be returned to the client per connection check, rather than
about 110k per status check.
Adding a more detail print functions, that assist/make it easier for
creating SNMP scripts.
They are only supported when you have NET-SNMP.
Need to be documented still though.
Adding SNMPv3 support.
This splits the old php_snmp function into an argument parsing part and
a data retrieval part. The data retrieval part is reused for SNMPv3.
SNMPv3 has only a different argument parsing.
THe code is not extremely beautiful, but having NET-SNMP and
UCD-SNMP support creates this. :-(
Documentation must still be done.
Marcus Boerger [Mon, 11 Nov 2002 20:52:09 +0000 (20:52 +0000)]
-make the flock() emulation a separate function named php_flock()
-use a define to make php_flock() available as flock() when necessary
# The emulated php_flock even works on NFS this will be used elsewhere.
Ilia Alshanetsky [Mon, 11 Nov 2002 16:12:56 +0000 (16:12 +0000)]
Added a flex version check. In the event required flex version is not
available, an advisory warning (similar to bison one) will be given to the
user. This is needed for systems such as Sun OS, where the existing lex
parser cannot be used to regenerate Zend/PHP lexical parsers.
Marcus Boerger [Mon, 11 Nov 2002 12:00:58 +0000 (12:00 +0000)]
-Precondiftion: no open files
-do not show warning from dba_open call for read during write
# Interesting is that on some systems read during write is permitted...
# I will change the tests as soon as i find a way to have this equal on all
# systems
Jason Greene [Mon, 11 Nov 2002 03:45:04 +0000 (03:45 +0000)]
add test function for log()
# It is difficult to test an approximation of an approximation, but I
# believe this test should be close enough to verify the function is working
# correctly.
Marcus Boerger [Sun, 10 Nov 2002 22:56:29 +0000 (22:56 +0000)]
Add $lock_flag which allows to change lock behaviour for both 'l' and 'd'
Remember gdbm has its own locking so 't' does not work here - maybe a
warning is needed for that, too.