Since the mysqlnd result set is arena allocated, we must not simply
free it, but rather call the appropriate `free_result` method.
Co-authored-by: Kamil Tekiela <tekiela246@gmail.com>
This feature was never implemented, and since the
beginning, using this constant with mysqlnd would
result in an error.
This feature was only available with libmysqlclient
which can no longer be used with mysqli.
There are no plans to implement it in the future.
`gettimeofday()` is supported by PHP on Windows for ages; and generally
`localtime()` is supported on Windows for a long time. As such, there
is no need for the Windows specific formatting code.
However, the general Windows caveat regarding `time_t` applies, namely
that it is usually `__time64_t`, unless `_USE_32BIT_TIME_T` is declared,
what we do for 32bit architectures, in which case it is `__time32_t`.
Now, `struct timeval` is imported from WinSock2.h, where the members are
declared as long (i.e. 32bit on both x86 and x64). That means passing
a pointer to `tv_sec` to `localtime()` likely fails on x64, or at least
doesn't yield the desired result. Therefore, we assign `tv_sec` to an
appropriate `time_t` variable, and also make sure that the `time_buffer`
is zero-terminated even if the `localtime()` call still fails.
Quoted m4_normalize will expand and change its argument later in the
macro call when M4 is processing the *.m4 sources. Without quotes the
already normalized string is passed to the macro directly. In these
specific cases generated configure script is the same. This is more for
consistency to have this synced and not repeat the pattern too much
in the future when copy/pasting. Note, that many AC_* macros require
similar behavior already (for example, AC_CHECK_FUNCS.)
* Include from build dir first
This fixes out of tree builds by ensuring that configure artifacts are included
from the build dir.
Before, out of tree builds would preferably include files from the src dir, as
the include path was defined as follows (ignoring includes from ext/ and sapi/) :
-I$(top_builddir)/main
-I$(top_srcdir)
-I$(top_builddir)/TSRM
-I$(top_builddir)/Zend
-I$(top_srcdir)/main
-I$(top_srcdir)/Zend
-I$(top_srcdir)/TSRM
-I$(top_builddir)/
As a result, an out of tree build would include configure artifacts such as
`main/php_config.h` from the src dir.
After this change, the include path is defined as follows:
-I$(top_builddir)/main
-I$(top_builddir)
-I$(top_srcdir)/main
-I$(top_srcdir)
-I$(top_builddir)/TSRM
-I$(top_builddir)/Zend
-I$(top_srcdir)/Zend
-I$(top_srcdir)/TSRM
* Fix extension include path for out of tree builds
* Include config.h with the brackets form
`#include "config.h"` searches in the directory containing the including-file
before any other include path. This can include the wrong config.h when building
out of tree and a config.h exists in the source tree.
Using `#include <config.h>` uses exclusively the include path, and gives
priority to the build dir.
The m4_normalize() can simplify the list argument. It removes leading
and trailing spaces and tabs, sequences of backslash-then-newline, and
replaces multiple spaces, tabs, and newlines with a single space.
This enables the zlib library (https://zlib.net/) from a single place to
match the minimum required version across the php-src. This provides a
possible simpler version bump in the future. Macro's 2nd and 3rd
arguments can pass additional actions whether zlib library is found or
not for the possible future adjustments in the ext/standard where also
zlib might be required.
Support for pkg-config was introduced in 1.2.3.1.
The minimum zlib version has been bumped to 1.2.3.1
* Bump minimum zlib version to 1.2.11
This is aligned with CentOS 8, which has 1.2.11 in the default packages.
* [skip ci] Move zlib version change to UPGRADING
This matches the OpenSSL style version change notice already done in
this file.
This partially backports that PR to stable branches as it has been in master
without reported problems so far.
It's only a partial backport because the stable branches don't have the
ZTS persistent resource fix that would fix shutdown crashes, i.e. the
code change in mysqlnd_vio's close_stream is not backported.
This is fully fixed on master.
Closes GH-14324.
Closes GH-10599.
This is used in Makefile when building mysqlnd as shared to get zlib and
crypto libraries linked in the mysqlnd.so (-lz -lcrypto). For static
build these are in the resulting binary like before.
This is a leftover from the refactoring of the --with-openssl-dir option
that once accepted the path to OpenSSL but wasn't renamed back then.
Instead of --with-openssl-dir, SSL support in ext/ftp and ext/mysqlnd
can be enabled implicitly when building with ext/openssl enabled
(--with-openssl) or explicitly by using new separate configure options
--with-ftp-ssl and --with-mysqlnd-ssl.
The code originally posted in GH-10599 triggers the bug for
non-persistent connections, but changing the host to `p:.` reveals that
there is also a crash bug for persistent connections.
What happens is that the persistent network stream resource gets freed, yet
stays inside EG(persistent_list). This causes a crash on shutdown when the
persistent list is getting cleared, as the engine will try to free the network
stream again.
The code in close_stream gets confused between persistent vs
non-persistent allocations when EG(active) is false.
This code was introduced in c3019a1 to fix crashes when the persistent list
gets destroyed before module shutdown is called. This is indeed a potential
problem that was fixed on the master branch in 5941cda.
This fixes the crash reason of GH-10599.
The x86-32 build uses a fast path, but when I disabled the fast path I
got the following compile error with -Werror:
mysqlnd_portability.h:221:95: error: right shift count >= width of type [-Werror=shift-count-overflow]
For 32-bit platforms that don't have the fast path, this can cause
undefined behaviour at runtime. Some CPUs just mask the shift amount
so in those cases you even get wrong results.
This is not always found during the build because -Werror is off by
default.
Almost every character set can be given a number N such that a multibyte
sequence starts with a byte higher than that number N. This allows us to
skip a lot of work. To ensure the correctness of this, a sanity check is
implemented that exhaustively tries every 4-byte sequence for every
character set and checks for consistency issues.
This finally gives:
Time (mean ± σ): 120.2 ms ± 1.2 ms [User: 116.9 ms, System: 2.8 ms]
Range (min … max): 118.0 ms … 122.9 ms 24 runs
We allocate twice the input length, and every input character results in
either 1 or 2 output bytes, so we cannot overflow.
By using an enum, and a switch table (which will be efficiently compiled
into a jump table), we can avoid the pessimistic code generation of the
indirect calls.
With this I get the following runtime for the test script in GH-13466 on
my i7-4790, which is around 1.25x faster.
Time (mean ± σ): 250.9 ms ± 1.6 ms [User: 248.4 ms, System: 2.0 ms]
Range (min … max): 248.9 ms … 254.4 ms 11 runs
The character set ID included in the handshake data at the time of connection
actually only includes the lower 8 bits of the ID, so if try to use this to specify
a character set, the corresponding character set may not exist.
In case of an invalid character set, the default character set is now used
without an error.
Fixes#13452Closes#13470
^_^