From an engine perspective, named parameters mainly add three
concepts:
* The SEND_* opcodes now accept a CONST op2, which is the
argument name. For now, it is looked up by linear scan and
runtime cached.
* This may leave UNDEF arguments on the stack. To avoid having
to deal with them in other places, a CHECK_UNDEF_ARGS opcode
is used to either replace them with defaults, or error.
* For variadic functions, EX(extra_named_params) are collected
and need to be freed based on ZEND_CALL_HAS_EXTRA_NAMED_PARAMS.
RFC: https://wiki.php.net/rfc/named_params
Closes GH-5357.
Even if that header file is available, we better consider it private,
and don't include it. The information about whether SSL support is
enabled is now missing (`USE_(OPEN)SSL`), and it seems there is no
alternative way to get it (`PQinitSSL()` is always defined), so we
remove it from the PHP info. Furthermore, the `PG_VERSION` and
`PG_VERSION_STR` macros are no longer available, but as of libpq 9.1
there is `PQlibVersion()` which allows us to construct `PG_VERSION` in
a most likely backwards compatible manner. The additional information
available through `PG_VERSION_STR` is lost, though, so we define
`PGSQL_LIBPQ_VERSION_STR` basically as alias of `PGSQL_LIBPQ_VERSION`,
and deprecate it right away.
Since we are now requiring at least libpq 9.1, we can remove some
further compatibility code and additional checks.
Regarding the raised requirements: official support for PostGreSQL 9.0
ended on 2015-10-08, and even CentOS 7 already has PostGreSQL 9.2, so
this is not supposed to be too much of an issue.
We can safely assume that users have at the very least libpq 7.4, for
which official support ended on 2010-10-01; even CentOS 6 has 8.4 now.
It is also noteworthy that PDO_PGSQL already requires libpq 7.4 or
later.
We have to actually determine the proper `SIZEOF_OFF_T`.
Interestingly, it is `4` on Windows x64.
We also have to prevent the redefinition in pg_config.h. The clean
solution would likely be to not include pg_config.h at all, but that's
out of scope for BC reasons for now.
Instead add RETURN_COPY(_VALUE) macros will the expected behavior.
RETURN_ZVAL doesn't make any sense since PHP 7, but has stuck
around, probably because the alternative was to write directly to
the return_value variable.
The php_stream_read() and php_stream_write() functions now return
an ssize_t value, with negative results indicating failure. Functions
like fread() and fwrite() will return false in that case.
As a special case, EWOULDBLOCK and EAGAIN on non-blocking streams
should not be regarded as error conditions, and be reported as
successful zero-length reads/writes instead. The handling of EINTR
remains unclear and is internally inconsistent (e.g. some code-paths
will automatically retry on EINTR, while some won't).
I'm landing this now to make sure the stream wrapper ops API changes
make it into 7.4 -- however, if the user-facing changes turn out to
be problematic we have the option of clamping negative returns to
zero in php_stream_read() and php_stream_write() to restore the
old behavior in a relatively non-intrusive manner.
RFC: https://wiki.php.net/rfc/tostring_exceptions
And convert some object to string conversion related recoverable
fatal errors into Error exceptions.
Improve exception safety of internal code performing string
conversions.
This is our most common intermittent failure. I don't understand
why it happens and can't reproduce it, but defining an explicit
sort order seems worth a shot...
