Fix intermittent test failures of windows_mb_path tests
Some of these tests create, use and later remove the same folder, so if
these are run in parallel, they may fail due to race conditions[1]. As
quick fix we add appropriate CONFLICTS clauses to prevent parallel
execution of the respective test groups.
Nikita Popov [Wed, 4 Mar 2020 09:10:36 +0000 (10:10 +0100)]
Implement interfaces after all methods available
The place where interface implementation handlers is called is
currently ill-defined: If the class implements interfaces itself,
the handlers for both the parent interfaces and the new interfaces
will be called after all methods are registered (post trait use).
If the class does not implement interfaces, then the parent
interface handlers are called early during inheritance (before
methods are inherited).
This commit moves the calls to always occur after all methods are
available. For userland classes this will be post trait import,
at the time where interfaces get implemented (whether the class
itself defines additional interfaces or not). For internal classes
it will be at the end of inheritance, as internal class declarations
do not have proper finalization.
This allows us to simplify the logic for implementing the magic
Iterator / IteratorAggregate interfaces. In particularly we can
now also automatically detect whether an extension of
IteratorAggregate can safely reuse a custom get_iterator handler,
or whether it needs to switch to the userland mechanism. The
Iterator case continues to rely on ZEND_ACC_REUSE_GET_ITERATOR
for this purpose, as a wholesale replacement is not possible there.
Nikita Popov [Tue, 3 Mar 2020 13:21:33 +0000 (14:21 +0100)]
Resolve trait alias refers to earlier
Make sure all trait method references are converted to absolute
method references in advance. This regresses one error message
that I don't think is particularly valuable.
Nikita Popov [Mon, 2 Mar 2020 10:07:57 +0000 (11:07 +0100)]
Store aliased name of trait method
Currently, trait methods are aliased will continue to use the
original function name. In a few places in the codebase, we will
try to look up the actual method name instead. However, this does
not work if an aliased method is used indirectly
(https://bugs.php.net/bug.php?id=69180).
I think it would be better to instead actually change the method
name to the alias. This is in principle easy: We have to allow
function_name to be changed even if op array is otherwise shared
(similar to static_variables). This means we need to addref/release
the function_name separately, but I don't think there is a
performance concern here (especially as everything is usually
interned).
There is a bit of complication in opcache, where we need to make
sure that the function name is released the correct number of times
(interning may overwrite the name in the original op_array, but we
need to release it as many times as the op_array is shared).
Fix #79013: Content-Length missing when posting a curlFile with curl
Unfortunately, some Webservers (e.g. IIS) do not implement the (F)CGI
specifications correctly wrt. chunked uploads (i.e. Transfer-encoding:
chunked), but instead pass -1 as CONTENT_LENGTH to the CGI
application. However, our (F)CFI SAPIs (i.e. cgi and cgi-fcgi) do not
support this.
Therefore we try to retrieve the stream size in advance and pass it to
`curl_mime_data_cb()` to prevent libcurl from doing chunked uploads.
This is basically the same approach that `curl_mime_filedata()`
implements, except that we are keeping already opened streams open for
the `read_cb()`.
Releasing the `com_dotnet_istream_wrapper` in `istream_destructor()` is
pointless, since `istream_destructor()` is only called when the
resource is going to be released. This recursion is not a real issue,
though, since the resource is never exposed to userland, and has at
most refcount 1, so due to well defined unsigned integer underflow, it
never is released twice. However, returning early in this case causes
a memory leak which needs to be fixed.
Nikita Popov [Fri, 28 Feb 2020 16:19:37 +0000 (17:19 +0100)]
Fixed bug #77325
Make ReflectionClassConstant->class the declaring class, not the
class on which the constant was fetched. This matches the behavior
for properties and methods.
Fix #79311: enchant_dict_suggest() fails on big endian architecture
For obvious reasons, we must not assign a `size_t` value to an `int`
variable using memcpy(). However, there is actually no need for the
intermediate `n_sugg_st` here, if we use the proper types in the first
place.
A regression test is not necessary, because dict_suggest.phpt already
exhibits the erroneous behavior on big endian architectures.
Nikita Popov [Thu, 27 Feb 2020 09:33:35 +0000 (10:33 +0100)]
Remove mergesort implementation
php_mergesort() isn't being used for anything, and hasn't been for
a long time. Even if we wanted to use a stable sort, this isn't
the implementation we'd use...
projects / php / log