We need to allocate buffers for the file mapping names which are large
enough for all potential keys (`key_t` is defined as `int` on Windows).
Regarding the test: it's probably never a good idea to use hard-coded
keys (should always use `ftok()` instead), but to reliably reproduce
this Windows specific issue we need to, and it shouldn't be an issue on
that OS.
Closes GH-7448.
If we assemble a zend_string manually, we need to end it with a NUL
byte ourselves.
We also fix the size calculation for that zend_string; there is no need
for the extra byte for each part, and we don't have to multiply by two,
since we're using DnsQuery_A(), not DnsQuery_W () (in which case we
would have to do the character set conversion, anyway). This avoids
over-allocation, and the need to explicitly set the string length.
Finally, we use the proper access macro for zend_strings.
Closes GH-7427.
We port the upstream fix[1], and also revert commit a3383ac3d7[2] which
is now obsolete, and also not part of libgd. Especially the change to
gd.png.c was at best a half-baked optimization.
[1] <a24e96f019>
[2] <a3383ac3d7>
Closes GH-7402.
We must avoid integer overflows in memory allocations, so we introduce
an additional check in the VM, and bail out in the rare case of an
overflow.
Closes GH-7381.
We need to run the full `php_cli_server_request_shutdown()` in case of
failing `php_cli_server_request_startup()`.
Patch contributed by @cataphract.
Closes GH-7322.
The fix for bug #73151[1] cured the symptoms, but not the root cause,
namely xmlParse() must not be called recursively. Since that bugfix
also messed up the error handling, we basically revert it (but also
simplify the return), and then prevent calling the parser recursively.
[1] <f2a8a8c068>
Co-authored-by: Nikita Popov <nikita.ppv@gmail.com>
Closes GH-7363.
Add a new interned string handler that fetches an interned string
if it exists, but does not create one if it does not (and instead
returns a non-interned string).
This fixes bug #81142, by preventing the creating of new interned
strings for unserialized array keys.
Closes GH-7360.
S->cols is already freed in the statement destructor and since
caa710037e the column data is only
populated on the first execute() which means that on subsequent
execute()s after closeCursor was called, all meta-data for column types
was removed and never restored
Closes GH-7355.
The ascii to wchar was reporting errors using conv_illegal_output,
while it should have been using WCSGROUP_THROUGH. Effectively that
replaced illegal characters with '?' for the purpose of
identification.
The stream position is not related to the buffer, and needs to be
updated for non-seekable streams as well. The erroneous condition
around the position update is a relict of an old commit[1].
The unexpected test expectation is due to bug #81345.
[1] <088e2692c3>
Closes GH-7356.
When flushing the stream filters actually causes data to be written to
the stream, we need to update its position, because that is not done by
the streams' write methods.
Closes GH-7354.
Fixes among others:
. Bug #79580 (date_create_from_format misses leap year).
. Bug #80974 (Wrong diff between 2 dates in different timezones).
. Bug #81097 (DateTimeZone silently falls back to UTC when providing an offset with seconds).
. Bug #81273 (Date interval calculation not correct).
The recent fix for bug 52093 is not compatible with LibreSSL ≥ 2.7.0,
which we recognize as mostly OpenSSL 1.1.0 compatible, but they still
do not support `ASN1_INTEGER_set_int64()`.
Closes GH-7339.
While our HTTP parser supports upgrade requests, the code using it does
not. Since upgrade requests are only valid for HTTP/1.1 and we neither
support any higher version, nor HTTPS yet, we do not exit early in case
of such requests, i.e. we ignore them, what is allowed by the specs.
We keep the supporting code in case we can meaningfully support upgrade
requests in the future.
Closes GH-7316.
If `P->len` is negative (not only when it is `SQL_NULL_DATA`), we must
not go on, because the following code can't deal with that. This means
that the output parameter will be set to `NULL` without any indication
what went wrong, but it's still better than crashing.
Closes GH-7295.
While parent:: should inherit the called scope, it should only do
so if it is compatible. If there is no called scope, or it is not
a subtype of the scope, we should fall back to the scope.
Use ZEND_MM_ALIGNED_SIZE for the extra size information.
I don't have a relevant system to test, but this should fix the
issue as long as required alignment is detected correctly.
When running without opcache, static_members_table is shared with
default_static_members_table. This is visible in reflection output,
because ReflectionProperty::getDefaultValue() will return the
current value, rather than the default value.
Address this by never sharing the table, which matches the behavior
we already see under opcache.
Fixes bug #80821.
Closes GH-7299.
Trying to allocate a `zend_string` with a length only slighty smaller
than `SIZE_MAX` causes an integer overflow, so callers may need to
check that explicitly. To make that easy in a portable way, we
introduce `ZSTR_MAX_LEN`.
Closes GH-7294.
This prevents serialization and unserialization of a class and its
children in a way that does not depend on the zend_class_serialize_deny
and zend_class_unserialize_deny handlers that will be going away
in PHP 9 together with the Serializable interface.
In stubs, `@not-serializable` can be used to set this flag.
This patch only uses the new flag for a handful of Zend classes,
converting the remainder is left for later.
Closes GH-7249.
Fixes bug #81111.
Adding two `zend_long`s may overflow, and casting `size_t` to
`zend_long` may truncate; we can avoid this here by enforcing unsigned
arithmetic.
Closes GH-7240.
Encoding a negative zero as `-0` is likely to loose the sign when
decoding (at least it does with `json_decode()`). Therefore, we encode
it as if `JSON_PRESERVE_ZERO_FRACTION` was specified, i.e. as `-0.0`.
Closes GH-7234.
Trimming a potentially over-allocated string appears to be reasonable,
so we drop the condition altogether.
We also re-allocate twice the size needed in the first place, and not
roughly tripple the size.
Closes GH-7231.
Implement printing for ReflectionAttribute. Attributes aren't
printed as part of reflection output for other structures (classes
etc) yet.
Closes GH-6117.
We must not assume that `usearch_last()` gives the proper result for
negative offsets. Instead we'd need to continue to search backwards
(`usearch_previous`) until we find a proper match. However, apparently
searching backwards is broken, so we work around by searching forward
from the start of the string until we pass the `offset_pos`, and then
use the previous result.
Closes GH-7189.
TMPVAR operands are destroyed using zval_ptr_dtor_nogc(), because
they usually cannot contain cycles. However, there are some rare
exceptions where this is possible, e.g. unserialize() return value.
In such cases we rely on the producing code to root the value. If
a GC run occurs between the rooting and consumption of the value,
we would end up leaking it. To avoid this, root all live TMPVAR
values after a GC run.
Closes GH-7210.
The analysis in the bug report wasn't correct (at least not in
this case -- there may still be a more general problem here),
the issue was that write_property returned the original variable_ptr
rather than the zend_assign_to_variable() return value, which will
DEREF the variable before overwriting it.
The built-in Webserver's `on_path`, `on_query_string` and `on_url`
callbacks may be called multiple times from the parser; we must not
simply replace the old values, but need to concatenate the new values
instead.
This appears to be tricky for `on_path` due to the path normalization,
so we fail if the function is called again.
The built-in Webserver logs errors during request parsing to stderr,
but this is ignored by the php_cli_server framework, and apparently the
Webserver does not send a resonse at all in such cases (instead of an
4xx). Thus we can only check that a request with an overly long path
fails.
Closes GH-7207.
When the time limit for a script is changed, when the script ends,
its INI value will be reset. This calls the event handler for the
timeout change, which will unset then reset the timeout. However,
this is done even if the script is done executing, and say, the CGI
or CLI web server process is idle.
This is probably incorrect, but isn't a problem on most platforms,
because PHP uses a timer that only ticks when the process is active
(that is, executing code). Since when it's idle, it's blocking on
listen/read, it won't tick because nothing executes. However, on
platforms where only the real-time timer is supported, (Cygwin/PASE)
it ticks regardless of if PHP is even executing. This means that the
idle processes are subject to timeouts from the INI reset on script
end.
This makes it so the timer is never set if the state is deactivating.
Testing with the CLI web server indicates the timer no longer
spuriously activates under PASE.
Closes GH-6683.
GMT+00:00 is recognized by ICU, and is normalized to GMT. There are no
issues when GMT+00:00 is passed to `IntlTimeZone::createTimeZone()`,
but passing it to IntlDateFormatter::__construct() causes a failure,
since there is an additional check regarding the validity. While
checking the validity of the result of `TimeZone::createTimeZone()`[1]
is a good idea, comparing the IDs is overly restrictive. Instead we
just check that the timezone is supported by ICU.
[1] <https://unicode-org.github.io/icu-docs/apidoc/dev/icu4c/classicu_1_1TimeZone.html#a35da0507b62754ffe5d8d59c19775cdb>
Closes GH-7190.
Adds support for the following options beginning with version 7.71.0:
CURLOPT_ISSUERCERT_BLOB
CURLOPT_PROXY_ISSUERCERT
CURLOPT_PROXY_ISSUERCERT_BLOB
CURLOPT_PROXY_SSLCERT_BLOB
CURLOPT_PROXY_SSLKEY_BLOB
CURLOPT_SSLCERT_BLOB
CURLOPT_SSLKEY_BLOB
Closes GH-7194.
When mapping the file, we need to pass the proper `dwFileOffsetHigh`
instead of `0`.
Co-authored-by: Nikita Popov <nikita.ppv@gmail.com>
Closes GH-7158.
This reverts commit a3e55286a6.
This commit broke the build for OSX before Sierra.
The apple documentation states the equivalent uses raw uptime, and
the source code of the non portable function says raw uptime would
call the same routine we were calling before this commit.
Since the correct version of this commit - according to docs -
would bring no value whatsoever, and this version breaks the build
there doesn't seem any sense in making this change.
We need to look for the right file. To avoid total confusion, we also
rename the variables; however, the right solution would have been to
drop the version suffixes, since they don't help (major + minor maybe,
but not major only).
When doing a non finishing flush, BZ2_bzCompress() returns BZ_FLUSH_OK
(not BZ_FINISH_OK) what requires us to do further flushes right away.
We also refactor the while-loop as do-loop.
Closes GH-7113.
When the memory limit is reduced using an `ini_set("memory_limit", ..)`
below the currently allocated memory, the out-of-memory check overflowed.
Instead of implementing additional checks during allocation,
`zend_set_memory_limit()` now validates the new memory limit. When
below the current memory usage the ini_set call will fail and throw
a warning.
This is part of GH-7040.
This is not guaranteed to work, since the actual server name may only
be given as SAN. Since we're doing the peer verification later anyway
(using the respective context options as appropriate), there is no need
to even supply a server name when verifying against the Windows cert
store.
Closes GH-7060.
For concatenation, the in-place variant can be much more efficient,
because it will reallocate the string in-place. Special-case the
typed property compound assignment code for the case where we
concatenate to a string, in which case we know that the result
will also be a string, and we don't need the type check anyway.
When row data split across multiple packets, allocate a temporary
buffer that can be reallocated, and only copy into the row buffer
pool arena once we know the final size. This avoids quadratic
memory usage for very large results.
(cherry picked from commit 1fc4c89214)
If ZTS is enabled, this can cause cwd_globals_ctor() to be called
multiple times, each with a freshly allocated virtual_cwd_globals
instance. At shutdown time however, cwd_globals_dtor() will call
realpath_cache_clean(), which then possibly cleans up the same
realpath_cache instance more than once. Using AddressSanitzer, this
shows up as a heap use-after-free.
To avoid this, add a helper function to do the actual work on one
instance of a realpath_cache, and call it both from cwd_globals_dtor()
and realpath_cache_clean(). The former uses the virtual_cwd_globals
parameter passed in via the destructor, the latter uses the CWDG()
macro.
Now that we properly dereference references of the superglobals. we
also need to dereference contained references to avoid to string
conversion.
Closes GH-7014.
Before the zval -> zend_object migration, this code used macros
like FORMATTER_METHOD_FETCH_OBJECT_NO_CHECK, which internally
clear the error. Now that they are no longer used, we need to
manually clear the error.
Literal compaction was incorrectly assuming that literals with
the same base literal and the same number of related literals
would be equal. Maybe that was the case historically, but at
least it isn't true in PHP 8, where FETCH_CONSTANT and INIT_METHOD
have distinct literals at the second position.
Fix this by making the cache key a concatenation of all literals,
rather than just the base literal. We still distinguish the number
of related literals based on a bias added to the string hash.