| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Implicitly constructing a value from a ReturnedValue muddies the
responsibility for ensuring that the gc can find the object.
With this change, we disable the implicit conversion. The expectation
for lifetime management is now:
- If a ReturnedValue is stored on the C++ stack, it must be put into a
QV4::Scoped class (or there should be a comment why not doing so is
safe). Passing a ReturnedValue to a function should no longer be
possible, unless the function takes a ReturnedValue, in which case the
expectation is that it stores the value in a place where it can be
seen by the gc, before doing anything that could trigger a gc run.
Using Value::fromReturnedValue can still be used to pass a Value on,
but in that case, the expectation is that there is a comment which
explains why this is safe.
- If a QV4::Value is obtained from a function call, it ought to be
stored in a ScopedValue, too. We currently can't enforce this easily,
so this should be checked during code review. A possible way forward
would be to disallow returning Values, but that would be a larger
change, and is deferred to the future.
- If a functions has a QV4::Value parameter, it's the callers'
responsibilty to ensure that the gc can find it.
Pick-to: 6.9 6.8 6.5
Fixes: QTBUG-131961
Change-Id: Iea055589d35a5f1ac36fe376d4389eb81de87961
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
|
| |
|
|
|
|
|
|
|
|
|
| |
We don't need the type arguments anymore. Change the functions to not
accept them and the code generator to not generate them.
This involves remembering the "isShadowable" flag until we potentially
have to re-initialize a lookup for a write back operation.
Change-Id: I036f87de20061ae9f2c7588a41bcdca0cd72df02
Reviewed-by: Olivier De Cannière <olivier.decanniere@qt.io>
|
| |
|
|
|
|
|
|
|
| |
It's currently impossible to generate code that uses them. Furthermore
the API is quite wrong in that it needlessly takes all types as part of
the call rather than the initialization.
Change-Id: Ic452cb68f43410d9ebafd29d55a4d02273d95b9b
Reviewed-by: Semih Yavuz <semih.yavuz@qt.io>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This allows us to create value types with invokable copy ctors from
JavaScript objects that describe their properties. That way we now have
a full replacement for the Qt.foo() methods.
As a side effect, we support retrieval of enums for certain kinds of
broken value types now, if prefixed with a namespace.
Fixes: QTBUG-124634
Change-Id: If2a3c59d647e211ef5a0cd1ddee04b409d9ea5f3
Reviewed-by: Olivier De Cannière <olivier.decanniere@qt.io>
|
| |
|
|
|
|
|
|
| |
Cache the metatype in the lookup rather than re-calculating it in every
call.
Change-Id: I5fde45a05a3d515466d9cf360bf51d7a8c9cb816
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
While the C++ standard says that different functions need to have unique
addresses, some compilers have take substantial liberties with that
rule. This means we can't actually rely on the addresses of our
different lookup functions to differ and therefore we cannot use them as
discriminator.
Introduce an enumeration for all the different lookups and use that
instead.
Now we can also drop all the purely redirecting methods we've introduced
just to have different addresses.
Change-Id: Ifa68c27c0d2fef4084893a19227dab21bd948dfd
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
| |
|
|
|
|
|
|
|
|
| |
Codechecker complains about ambiguous 1-letter names, and especially 'l'
vs 'I'.
Change-Id: Ia32c2a8eb04956bb10e7e49e2085b724e8feb749
Reviewed-by: Olivier De Cannière <olivier.decanniere@qt.io>
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that resolving the types can be more expensive, especially for
composite types, we should be careful not to do it too often. Most of
the cases where we have to resolve types are for lookups. Most of the
lookups only resolve types in their "init" step, which ideally is only
performed once in the application life time.
However, so far calls had to pass the types of arguments and return
values in the actual "lookup" step, which causes the resolution to
happen on every call.
This change moves those type resolutions to the init step. We can do
this because:
1. Regular typed method calls are calls to a specific overload with
specific types. The method itself already has the list of types and
we can just remember which one that is. Then we don't need to pass
the types.
2. Calls to shadowable methods are all-QVariant typed. The only thing
we need to know is the number of arguments. Then we can construct
the list of types in the lookup itself.
We can remember which one of those we're dealing with by adding further
"AsVariant" lookup functions. For the case of non-property-cached
regular methods we also need a new "Fallback" lookup like we already
have it for properties.
Change-Id: I74a3729131d6a5ea0ad79e276965a5167cd609be
Reviewed-by: Olivier De Cannière <olivier.decanniere@qt.io>
|
| |
|
|
|
|
|
|
|
|
|
| |
The old function might still be callde from code compiled against Qt <
6.6. Fix this by providing the old overload in removde_api: It is
implemented by calling the new overload, and then converting the result
to an int. Amends f10630aa035ae1ef4ad75be1107ab77fea0c7e29.
Pick-to: 6.6
Change-Id: I373460af2a66c069bfc0b57a9bb169d705d2a6b9
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
|
| |
|
|
|
|
| |
Pick-to: 6.5
Change-Id: I7f700d64694c8651769841a0109d32e8f9a839b2
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
|
|
Add explicit where appropriate, and use more elegant constructs in
inline functions. Introduce removed_api.cpp for
QJSEngine::create(int, const void *).
Pick-to: 6.5
Change-Id: Ie54b0494fe3c5567f8a5ca361c3a583de3d97dd5
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
|
