The header klib is supposed to only contain the public abi of the module
similar to jvm-abi-gen. It is intended to be used as a dependency for other
klib compilations instead of the full klib for compilation avoidance.
^KT-60807
Previous attempt was at 6807ed6642, but it was reverted in 7771e5914d
(see that commit for the explanation). Now it should be fine to remove
these version requirements.
No issue reported because it's mostly not a user-visible change. The
main effect is that it reduces differences in metadata between K1 and
K2, namely it fixes the ValueClass.kt kotlinp test.
Language feature InlineClasses is enabled since 1.3. The current lowest
supported language version is 1.4, so any compiler that can read
binaries produced by the current compiler also supports inline classes,
which means that the version requirement is not needed anymore.
No issue reported because it's mostly not a user-visible change. The
main effect is that it reduces differences in metadata between K1 and
K2, because K2 never supported writing this version requirement properly
due to the TODO in `hasInlineClassTypesInSignature`.
Sorting invokes descriptor renderer, so the fewer descriptors we are
sorting, the better. This helps to save a few percent of compilation
time in a synthetic project from KT-52326.
Always serialize type table (and version requirement table) after
everything else, to prevent bugs like KT-51446 where inline class
underlying type was added to the type table after it has already been
serialized.
#KT-51446 Fixed
This reverts commit 6807ed6642.
The reason is that Kotlin compilers until and including version 1.6.10
have code that detects obsolete coroutines via these version
requirements (see `versionAndReleaseCoroutinesMismatch`). Since Kotlin
1.6.10 can read metadata of version 1.7, the earliest we can drop these
version requirements is Kotlin 1.8.0.
Coroutines are a stable feature since 1.3. Version requirement on
suspend functions, or functions mentioning suspend function types in the
signature, was needed to prevent older compilers from reading metadata
that they can't properly use. It's not needed anymore because a newer
metadata version prevents older compilers from reading any metadata
altogether.
Also, computing isSuspendOrHasSuspendTypesInSignature took ~0.5% of
backend time on compilation of intellij (related to KT-48233).
Anonymous types are not approximated by frontend for private declarations.
Class IDs for such types are replaced by StringTable before being written
to metadata. JVM string table mangles such types and keeps them generic
for reflection. For other purposes the types are replaced with `Any`.
Type arguments of the replaced type should be ignored in the latter case.
Otherwise decompiled text builder crashes on an attempt to restore `Any`
type with non-zero number of type arguments.
The code pretending to replace a type with its first supertype was dropped
from ApproximatingStringTable for two reasons:
- the first type from `getAllSuperClassifiers` is the original type itself
which doens't provide a ClassId for anonymous type, so it was a noop
- tracking potential type arguments of the first anonymous type's supertype
would be a complication with almost no practical value (types in decompiled
text would be slightly closer to the real type of a private declaration).
^KT-46393 Fixed
This will be used in the compiler starting from 1.5.20 instead of the
currently used approach of looking for the single value parameter of the
primary constructor. The problem with the current approach is that
primary constructor can be private (since 1.4.30) and the property could
always be private. Relying on private declarations from metadata is
dangerous; for example lazy IR doesn't usually create stubs for private
declarations, and it didn't create stubs for private inline class
constructors before b5f9b1df, which led to the problem reported in
KT-44723.
Take local/anonymous classes and their inner classes into account.
Simplify approximation: use first available super classifier instead
of first super class. This approximation should only happen for
private declarations that were not previously approximated by frontend.
So basically the only requirement for the approximated types is to be
denotable. Note that this only works if the types are not used later.
JVM uses a different string table implementatin as it needs exact
types of private members for reflection.
^KT-20996 Fixed
instead of statically registering it.
Static registering can cause subtle errors when plugin implementation
(e.g. SerializationDescriptorPluginForKotlinxSerialization) is registered
from multiple classloaders: in multi-module with daemon compilation scenario
#KT-41857 Fixed
There is two methods added -- `or` and `and`
`or` is used by CFA for combining two kinds that came from different
edges of control flow graph
`and` is analog of `+` operator for invocation kinds
A mechanism that allows kotlinx.serialization plugin to preserve the
correct (program) order of properties after serializing/deserializing
descriptors to kotlin metadata, which is needed for correct and stable
json serialization of class hierarchies in incremental/multi-module scenario.
It uses protobuf extensions.
Replace it with a dependency on 'descriptors'.
Move the existing marker interface ContractProvider to 'descriptors',
and create a new marker interface DeserializedDescriptor.
Johan Bay