Primarily to avoid getting fatal errors from token_get_all().
Implemented using a magic E_ERROR token, which the lexer emits to
force a parser failure.
* PHP-5.6:
Prevent GC from changing zval or object 'color' before they are actually inserted into possible roots buffer.
Conflicts:
NEWS
Zend/zend_gc.c
* InternalClassClean:
Fixed test
Patch improvement:
Fixed indentation. Fixed comment style. Fixed commented out code.
Reverted change to function name and added note of why it is different from the class it is actually changing.
Made UConverter throw an exception if the constructor fails.
Fixed PDO constructor to not return null.
Fixed fileinfo behaviour.
Made Phar throw exception on bad constructor.
Converted intl extension to use IntlException in constructors.
Fixed SplFixedArray and tests.
Fixed ReflectionExtension and ReflectionProperty.
Fixed ReflectionFunction, ReflectionMethod and ReflectionParameter.
Fixed PDORow behaviour and message.
* The class names false, true and null are now reserved.
* The code dealing with reserved class names is now decoupled
from scalar type hint handling. It also includes self, parent,
and static, which are class names which were already reserved
previously.
* Reuse existing messages for reserved class names.
Fallout: class_alias() can no longer alias self, parent and static.
However this never really worked in the first place, as the test
which was testing this shows.
C:\> php -r "trait A { } trait A { }"
Will now properly print "Cannot redeclare trait A" instead of "Cannot redeclare class A" to make error messages a tiny bit clearer. Admittedly, a better solution can most likely be made by actually telling what the colliding object is a type of.
Internally this adds a new function:
zend_get_object_type()
ZEND_ADD_STRING/VAR/CHAR are replaced with ZEND_ROPE_INTI, ZEND_ROPE_ADD, ZEND_ROPE_END.
Instead of reallocation and copying string on each ZEND_ADD_STRING/VAR/CAHR, collect all the strings and then allocate and construct the resulting string once.
a) The SAME_FAKE_TYPE check should always occur on the dereferenced
value, otherwise we'll end up separating large referenced arrays.
b) For by-val return with an rc=1 reference, the SEPARATE_ZVAL
wouldn't do anything. Add a ZVAL_UNREF for this case.
I don't like this fix - there must be some better way to express
this.
This is allowed as per the return types RFC. The test for this
behavior happened to use an internal class without arginfo, which
is why this was not properly implemented.
EX_PREV_USES_STRICT_TYPES() and family changed/renamed to fit with other macros
Optimized zend_verify_internal_arg_type() and family (they don't need "strict" argument anymore)
Standerd ZPP is called from VM only for weak type check or strict exception (int -> double)
Fixed ZEND_RECV_VARIADIC
Fixed ZEND_STRLEN
TODO: should we accept IS_NULL for non-nullable arguments?
* 'scalar_type_hints_v5' of github.com:ircmaxell/php-src: (54 commits)
Add check for maintaining reference all the way through both type and return values
Reduce the number of times that the zval needs to be separated in return type checking to those that are necessary
Add test to ensure namespaced code can't use scalar types as class names
Disallow relative namespace type declarations
Add support and tests for null constant default values. Refactor complex conditionals into an extracted function for clarity and code-reuse
Refactor as to not use call info, but add the flag to the op_array.
Fix severity issues with callbacks, start work porting ZEND_STRLEN opcode to work with strict mode, more refactoring to come
Fix C89 compatibility by moving a misplaced if statement
Refactor gotos into more elaborate ifs to eliminate goto failure
Fix return type separation with references. It now includes a check in the opcode handler and properly separates the value in both cases
Add error case with invalid type declaration default value
Add tests for using constants as defaults with scalar type declarations
Refactor error messages to only append defined in the event of rendering as a fatal error. Fix relative typehints to not work
Revert gd and mysqli to former calls for arg count errors. Fix first identified issues with zend_compile.c from nikic. Add tests for generator returns prior to first yield
Style cleanup, as well as fixing bug with missing argument for WRONG_PARAM_COUNT_WITH_RETVAL
Clean up API renames for zend_wrong_param_count to maintain BC, introducing a zend_wrong_param_count_ex() function when you know strict value
Fix executor issue with ignoring strict types, which cleans up a bunch of errors. Additionally fix the expected error of 2 unrelated tests that was caused by a change to the core error messages
Fix more tests, improve error messages in failure output for thrown exception
Refactor error implementation significantly to centralize error mode behavior. Add zend_internal_type_error() function
Refactor declare placement detection, allow encoding declares before type declares
...
There's always a possibility that the has will end up as equal to zero due to overflowing.
See the tests for strings whose hash overflow.
Yes, there's a check for whether the hash already was calculated... and strings with a hash equal to zero always need the hash recalculated, in one of 2^32 respectively 2^64 cases - which should not be a major issue.
Removed HashTable->arHash (reduced memory consumption). Now hash slots may be accessed using HT_HASH() macro.
Hash slotas are allocated together with Buckets (before them) and lay in reverse order from HashTable->arData base address (see comments in Zend/zend_types.h)
Indexes in hash table and conflict resolution chains (Z_NEXT) may be stored as indeces or offsets in bytes, depending on system (32 or 64-bit).
HashTable data filelds are reordered to keep the most useful for zend_hash_find() data in the same CPU cache line.
This is disabled by default yet, but may be enabled compiling zend_execute.c with -DHAVE_GCC_GLOBAL_REGS.
Only tested on Linux x86 and x86_64 with GCC 4.9.2.
```
use function N\S { foo, bar, etc };
use const N\S { FOO, BAR, ETC };
```
Related tests:
```
make test -j4 TESTS="-m Zend/tests/ns_0{88..92}*.phpt"
```
All namespace tests:
```
make test -j4 TESTS="Zend/tests/ns_*.phpt"
```
actually, maybe we should precalculate before calling zend_hash_appen
when we are not sure whether the hash is caclculated(prop_info->name).
but it looks a little ugly.. (also for zend_string_copy)
Generators all have their own vm_stack, but usually don't go really deep.
In most cases even most of the 4 KB will remain unused (in my tests average is rather 1 KB), but this was tested a straightforward implementation without a lot of abstraction layers.
That way I'm using a more conservative 4 KB stack size which really should be enough for, I think, at least 90% of the cases.
This was necessary to not immediately run out of memory with 95%+ usused vm_stack space: If you have a few thousands of Generators active, your application quickly hits memory_limit with a stack page size of 256 KB...
In addition, it's also a bit faster (70% less instructions for zend_vm_stack_new_page) due to emalloc() not having to allocate a whole new segment (segment size 256 KB). Also no mmap()/malloc() necessary.
See https://wiki.php.net/rfc/expectations
Squashed commit of the following:
commit 3f3651a787
Author: Dmitry Stogov <dmitry@zend.com>
Date: Mon Mar 2 11:56:33 2015 +0300
opcode 137 is used for ZEND_OP_DATA and con't be reused for ZEND_ASSERT_CHECK
commit ca8ecabf2a
Merge: 24328ac9dac923
Author: Dmitry Stogov <dmitry@zend.com>
Date: Mon Mar 2 10:49:23 2015 +0300
Merge branch 'master' into assert
* master:
Update NEWS
Fixed bug #69139 (Crash in gc_zval_possible_root on unserialize)
windows only test
Align entries format
Align entries format for 5.6.7
Align entries format for 5.5.23
Bump header year
Fixed bug #69144 (strtr not replacing with partly matching replace pairs)
Fixed test?
Revert mktime()/gmmktime()'s arginfo
Update NEWS
Fixed bug #69141 Missing arguments in reflection info for some builtin functions
Add NEWS entry
Remove useless date warning
Fix ARG_INFO for levenshtein
Fix ARG_INFO for levenshtein
fix dir separator in tests
Update NEWS
Fixed bug #69085 (SoapClient's __call() type confusion through unserialize()).
commit 24328ac03f
Merge: 021fd941cdee9a
Author: Dmitry Stogov <dmitry@zend.com>
Date: Fri Feb 27 15:57:13 2015 +0300
Merge branch 'master' into assert
* master:
Fixed C++ support
Fixed bug #69115 crash in mail
Reorder
Update NEWs
Fixed bug #69121 (Segfault in get_current_user when script owner is not in passwd with ZTS build)
Update News
Fixed bug #69125 (Array numeric string as key)
fix bug#68942's patch
Fixed ability to build unspecialized executor
Fixed bug #69124 (method name could not be used when by ref)
Fixed a bug that header value is not terminated by '\0' when accessed through getenv().
Fixed a bug that header value is not terminated by '\0' when accessed through getenv().
commit 021fd94ed1
Merge: 49963ebace1f82
Author: Dmitry Stogov <dmitry@zend.com>
Date: Thu Feb 26 11:26:03 2015 +0300
Merge branch 'master' into assert
* master: (59 commits)
Improved ASSIGN_DIM handler
Don't inline slow path
Revert a part committted by mistake
Fixed compilation warnings
Fixed a bug that header value is not terminated by '\0' when accessed through getenv().
better name
Improve fix for #69038
Update NEWs
Fixed bug #69108 ("Segmentation fault" when (de)serializing SplObjectStorage)
Added specialized versions of DO_FCALL handler: DO_ICALL - for internal functions DO_UCALL - for user functions DO_FCALL_BY_NAME - plain, most probably user, funcstions (not methods)
Use cache_slot offsets instead of indexes (simplify run-time instructions)
Split INIT_FCALL_BY_NAME inti INIT_FCALL_BY_NAME(CONST+STRING) and INIT_DYNAMIC_CALL(CONST-STRING|TMPVAR|CV)
Support list($a, $b) = $a
Avoid unnecassary check
Make zend_array_destroy() to free the corresponding zend_array
Eliminate check on the fast path
Make current() and key() receive argument by value.
Use Firebird default home folder, replace Interbase with Firebird
Updated NEWS
updated NEWS
...
Conflicts:
Zend/zend_vm_execute.h
Zend/zend_vm_opcodes.c
Zend/zend_vm_opcodes.h
commit 49963ebf9d
Merge: 07b1f926b77959
Author: Dmitry Stogov <dmitry@zend.com>
Date: Thu Feb 19 11:13:08 2015 +0300
Merge branch 'master' into assert
* master:
Implemented AST pretty-printer
update NEWS to match the actual stuff in 5.6.6
update NEWS to match the actual stuff in 5.5.22
update NEWS(add missing entry for the enchant fix, and reorder the entries a bit)
fix typo in bug#
update NEWS
fix email format
update NEWS
update 5.6.6 release date in NEWS
Fix bug #69033 (Request may get env. variables from previous requests if PHP works as FastCGI)
BFN
fix test
fix test
fix test
Fixed bug #65593 (Segfault when calling ob_start from output buffering callback)
Updated NEWS
add CVE
5.4.39 next
Fix associativity to match Perl
Blast off to space.
Conflicts:
Zend/zend_ast.c
commit 07b1f92ed6
Author: Dmitry Stogov <dmitry@zend.com>
Date: Wed Feb 18 23:06:32 2015 +0300
fixed pretty-printer (support for "elseif")
commit 5a976c8d85
Author: Dmitry Stogov <dmitry@zend.com>
Date: Wed Feb 18 19:50:08 2015 +0300
Fixed vaeious ptetty-printer issues
commit 69491e8e8e
Merge: 84731573ddc246
Author: Dmitry Stogov <dmitry@zend.com>
Date: Wed Feb 18 10:18:32 2015 +0300
Merge branch 'master' into assert
* master:
Set PHP_JSON_VERSION to 1.4.0
Remove unnecessary resource checks in openssl ext
JSON is now maintained
commit 8473157fbb
Author: Dmitry Stogov <dmitry@zend.com>
Date: Wed Feb 18 10:17:26 2015 +0300
Fixed typo and white spaces
commit 96de5ffc8d
Author: Dmitry Stogov <dmitry@zend.com>
Date: Wed Feb 18 00:28:39 2015 +0300
Fixed assert() in namesapaces
commit 5eba069c28
Merge: 4a2d9c0d428bf2
Author: Dmitry Stogov <dmitry@zend.com>
Date: Tue Feb 17 22:45:55 2015 +0300
Merge branch 'master' into assert
* master: (25 commits)
improve debugability in TS debug builds
More UPGRADING, in particular on foreach
Fixed bug #69038 (switch(SOMECONSTANT) misbehaves) for master
Replace var is introduced
abstain from using xmlCleanupParser
fix TS build
Fix bug #68942 (Use after free vulnerability in unserialize() with DateTimeZone)
update news
Fix bug #68942 (Use after free vulnerability in unserialize() with DateTimeZone)
Port for for bug #68552
Fix bug #68942 (Use after free vulnerability in unserialize() with DateTimeZone)
Update NEWS
Fixed bug #69038 (switch(SOMECONSTANT) misbehaves)
- BFN
Don't read the local php.ini when Generating Phar When building phar shared, you can end up loading a previous phar.so that isn't compatible with the php cli being used to generate Phar here.
- Fixed bug #67827 (broken detection of system crypt sha256/sha512 support)
Delete json outdated package.xml
made ZEND_TSRMLS_CACHE_* macros look like function calls
- Fix merge
- Fixed bug #67427 (SoapServer cannot handle large messages) patch by: brandt at docoloc dot de
...
commit 4a2d9c0953
Author: Dmitry Stogov <dmitry@zend.com>
Date: Tue Feb 17 22:45:10 2015 +0300
Implemented AST pretty-printer to capture expression passed to assert()
commit 7a059b66d5
Merge: 9973df73892eba
Author: Dmitry Stogov <dmitry@zend.com>
Date: Mon Feb 16 18:42:28 2015 +0300
Merge branch 'expect' of github.com:krakjoe/php-src into assert
* 'expect' of github.com:krakjoe/php-src:
import expect
Conflicts:
Zend/zend_compile.c
Zend/zend_execute_API.c
Zend/zend_globals.h
Zend/zend_vm_def.h
Zend/zend_vm_execute.h
Zend/zend_vm_opcodes.c
Zend/zend_vm_opcodes.h
ext/opcache/Optimizer/block_pass.c
ext/opcache/Optimizer/pass1_5.c
ext/standard/assert.c
ext/standard/tests/assert/assert_error3.phpt
commit 3892eba2bf
Author: krakjoe <joe.watkins@live.co.uk>
Date: Sun Feb 2 12:49:35 2014 +0000
import expect
