Rework rendering of kt-like dump and signatures dump in order to avoid
unstable blank line between declarations of the same level:
1. No blank line for the first declaration inside the member scope of
the class.
2. Always a single blank line between each two subsequent declarations
inside the member scope of the class.
1. Local declarations don't participate in IR-linkage, because they
can be referenced only inside the same body -> can be dropped
from IR text tests.
2. Mangled names for private declarations computed by descriptors/fir
are actually not used anywhere (they are recomputed by IR
immediately before serialization of IR). But sometimes such
mangled names diverge between K1 and K2 -> don't check them, but
always check mangled names computed by IR even for private
declarations.
3. Also: Drop DUMP_LOCAL_DECLARATION_SIGNATURES test directive.
^KT-57428 Obsolete
^KT-57430 Obsolete
^KT-57434 Obsolete
^KT-57778 Obsolete
^KT-57775 Obsolete
We plan to disable computing full mangled names of declarations in all
manglers except the IR mangler (see the subsequent commits).
From now on, in irText tests we dump only the following mangled names:
- Full mangled names computed using the IR mangler
- Signature mangled names computed using the Descriptor mangler
- Signature mangled names computed using the IR mangler
- Signature mangled names computed using the FIR mangler
Here by a full mangled name we mean the mangled name of a declaration
computed using the `MangleMode.FULL` mode. Those mangled names include
the mangled names of the declaration'a parents.
By a signature mangled name we mean the mangled name of a declaration
computed using the `MangleMode.SIGNATURE` mode.
These mangled names are used to compute an `IdSignature` for
the declaration, hence the name.
Previously, creating a declaration with Fir2IrCallableDeclarationsGenerator/
Fir2IrClassifiersGenerator didn't guarantee that this declaration will
be actually added to the list of parent class/file declarations, which
lead to situations when FIR2IR created some declarations in the air
(mostly fake-overrides)
In this commit we have a lot of change in test data. This was caused
by the way where we evaluate constants. We split constant evaluation
into two distinct parts: only necessary evaluations for `fir2ir`
(like const val and annotations) and optimizations for lowering.
Now we don't do all constant evaluation on `fir2ir`, but IR
dump is executed after this phase, so test data changed.
#KT-58923
This doesn't reduce the quality of tests, because the flags are still
printed for declarations themselves. We only omit them in references.
However, this makes the tests more compatible with non-JVM backends
(see KT-58605), because flags of referenced stdlib declarations may
differ among target platforms.
For type parameters of generic properties,
`DeclarationDescriptor#getContainingDeclaration` must return
the property descriptor instead of the accessor function descriptor.
^KT-57436 Fixed
The reason #1 for this feature is that we want to test IdSignatures
generated for declarations. Currently, there is no (easy) way to ensure
that a change in the signature building logic doesn't cause any breaking
changes wrt klibs.
Now, most IdSignatures include hashed mangled names in them, so even if
we catch a regression where the included hash changes, there would be no
way of knowing immediately what caused it, unless we'd also have mangled
names in the expectations.
The reason #2 is to test the manglers themselves. Currently, there are
no tests for them. They heavily duplicate each other, this is already
causing issues (see KT-57427) that would be very hard to catch without
these tests.
^KT-58238 Fixed
- Huge refactoring for IC
- Update hash combination logic
- Introduce value class for IC hashes
- Calc md5 directly by function IR
- Split IC logic by classes
- Move JsIrLinkerLoader into separate file
- CacheUpdateStatus is a sealed class
- Render TYPE_PARAMETER reified flag
^KT-51081 Fixed
^KT-51084 Fixed
Descriptors are already supposed to be sorted in scopes. The problem is
that rendering descriptors for sorting takes a lot of time (~1.5% of
total compilation time of intellij with JVM IR), and simple heuristics,
like comparing by names first, don't fully help with it.
#KT-48233
Also improve an error message in `IrType.erasedUpperBound`, which seems
like a frequent first place where the JVM IR backend crashes in case an
error type has made it past psi2ir.
This will help in diagnosing problems such as KT-45016.
Before this commit we initialized delegate fields in primary constructor,
that could provoke NPE in case delegate is used in initializer of
some property backing field.
Now we initialize delegate fields directly instead.
Pavel Kunyavskiy