Instead of checking the type manually. Peculiarily the zpp
message does not use "array or object", but that's a general issue.
And I guess it's good that object support is not being advertised.
It turns out that all of these places ultimately only accept arrays,
not objects, so we can use Z_ARRVAL_P everywhere.
Also add _deref in a few places where the hash lookup result is
directly type-checked.
Instead overload get_properties_for for a few specific cases such
as array casts. This resolves the issue where ArrayObject
get_properties may violate engine invariants in some cases.
This resolves the long-standing issue where var_dump
a DateTime (etc) object makes a number of additional properties
accessible, which may also change other behaviors as a side-effect.
Setting up an empty default handler is not only useless, but actually
harmful, since internal entity-references are not resolved anymore.
From the libexpat docs[1]:
| Setting the handler with this call has the side effect of
| turning off expansion of references to internally defined general
| entities. Instead these references are passed to the default
| handler.
[1] <https://www.xml.com/pub/1999/09/expat/reference.html#setdefhandler>
ICU 59 already requires C++11 by default. The minimum version required
by the core is 50, which is compiled with at least C++11 in many distros
as package defs tell. Headers for ICU versions between ICU 50 and 58 look
fine when included for C++11 compilation, the linking is thereof not affected.
The macro PHP_CXX_COMPILE_STDCXX is based on
https://www.gnu.org/software/autoconf-archive/ax_cxx_compile_stdcxx.html
The patch consists on allowing to read the corresponding switch in a user
defined variable instead of enforcing CXXFLAGS globally. That way, every
ext or SAPI can decide, which C++ standard is to be used. The
documentation is provided in the m4 file.
C++11 is already somewhat older standard, C++14 were better. However
issues with GCC < 5.0 and some other compilers are possibly to hit back.
Still there's some time to check for C++14 for ext/intl, too. Having said
that, C++11 in ext/intl and a mechanism to determine features is a good step
towards better C++ support.
Since there is no need to patch libsqlite3 for our purposes, and since
libsqlite3 ≥ 3.3.9 (which is our current requirement) is widely
available on distros, there is no reason anymore to bundle the library.
Besides removing the bundled libsqlite, and adapting the configuration
respectively, we also fix the use of the SQLITE_ENABLE_COLUMN_METADATA
compile time constant to detect whether sqlite3_column_table_name() is
available by a working feature detection (otherwise bug_42589.phpt
would fail). We also skip bug73068.phpt for libsqlite 3.11.0 to
3.14.1 which have a bug (<https://sqlite.org/src/info/ef360601>).
We also completely drop support for the obscure pdo_sqlite_external
extension (which could have been enabled on Windows only by passing
`--pdo-sqlite-external` to configure), since it is not needed anymore.
Furthermore, we remove references to the bundled libsqlite from
Makefile.gcov, CONTRIBUTING.md and README.REDIST.BINS.
run-tests.php assumes the contents of an ENV section are a simple list
of key-value-pairs, instead of evaluating the section contents as PHP
script which server-tests.php did. Since server-tests.php is gone, we
remove the superfluous and confusing scripting.
Nikita Popov