Directly fail unserialization when trying to acquire an r/R
reference to an UNDEF HT slot. Previously this left an UNDEF and
later deleted the index/key from the HT.
What actually caused the issue here is a combination of two
factors: First, the key deletion was performed using the hash API,
rather than the symtable API, such that the element was not actually
removed if it used an integral string key. Second, a subsequent
deletion operation, while collecting trailing UNDEF ranges, would
mark the element as available for reuse (leaving a corrupted HT
state with nNumOfElemnts > nNumUsed).
Fix this by failing early and dropping the deletion code.
* PHP-7.0:
Fix bug #73737 FPE when parsing a tag format
Fix bug #73773 - Seg fault when loading hostile phar
Fix bug #73825 - Heap out of bounds read on unserialize in finish_nested_data()
Fix bug #73768 - Memory corruption when loading hostile phar
Fix int overflows in phar (bug #73764)
* PHP-5.6:
Fix bug #73737 FPE when parsing a tag format
Fix bug #73773 - Seg fault when loading hostile phar
Fix bug #73825 - Heap out of bounds read on unserialize in finish_nested_data()
Fix bug #73768 - Memory corruption when loading hostile phar
Fix int overflows in phar (bug #73764)
* PHP-5.6.30:
Fix bug #73737 FPE when parsing a tag format
Fix bug #73773 - Seg fault when loading hostile phar
Fix bug #73825 - Heap out of bounds read on unserialize in finish_nested_data()
Fix bug #73768 - Memory corruption when loading hostile phar
Fix int overflows in phar (bug #73764)
If a (nested) unserialize() call fails, we remove all the values
that were inserted into var_hash during that call. This prevents
their use in other unserializations in the same context.
Don't call __destruct() on an unserialized object that has a
__wakeup() method if either
a) unserialization of its properties fails or
b) the __wakeup() call fails (e.g. by throwing).
This basically treats __wakeup() as a form of constructor and
aligns us with the usual behavior that if the constructor call
fails the destructor should not be called.
The security aspect here is that people use __wakeup() to prevent
unserialization of objects with dangerous __destruct() methods,
but this is ineffective if __destruct() can still be called while
__wakeup() was skipped.
The (UN)SERIALIZE_INIT/DESTROY macros now go through non-inlined
functions, so any changes to them will apply to extensions without
rebuilds.
Additionally, the (un)serialize_data structures are now no longer
exported.
This means that we are allowed to change these structures in patch
releases without breaking the ABI.
These are either in debug code (fix them), commented out (drop
them) or in dead compatibility macros (drop them).
One usage was in php_stream_get_from_zval(), which we have not used
since at least PHP 5.2 and, judging from the fact that nobody
complained about it causing compile errors in PHP 7, nobody else
uses it either, so drop it.
There are still remaining uses in mysqli embedded and odbc birdstep.
These probably need to be dropped outright.
