IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
In standalone builds, it could otherwise happen that the variable is
not set during the first CMake invocation and tests fail because they
are unable to locate the LLVM headers.
This effectively reverts commit 74472caaa9 ("[cling] Fixes issue in
DeclUnloader: do not unload templates intantiated in the PCH"), it's
not needed anymore, all tests pass and the snippet in the summary of
https://github.com/root-project/root/pull/4447 works, but it filters
too many declarations from the unloader.
The body of member functions of a templated class only gets instantiated
when the function is used. These `CXXMethodDecl` should not be deleted
from the AST; instead return them to the 'instantiation pending' state.
StaticVarCollector recursively visited descendants of a `FunctionDecl`
node to collect static local variables. However, these always appear
in the enclosing DeclContext.
This expands the handling introduced in commit 31e46a65f6 to include
the length modifiers 'short' and 'long' as mentioned back then. It
still doesn't handle the cases of multiple modifiers such as "unsigned
long long int" and "long int unsigned long" (being the same type).
Fixes https://github.com/root-project/root/issues/13288
Otherwise they are emitted as internal and we get double-construction
and -destruction on the same memory address due to the way we promote
internal declarations in KeepLocalGVPass.
According to upstream tests, the de-duplication doesn't work on Windows
(yet), but I think this problem is severe enough to fix it on the other
platforms sooner rather than later.
Fixes#13429
(For context, this is important for multi-dimensional constant arrays
as described in ROOT-7016 and tested in Cling's Interfaces/evaluate.C
test. But for reasons unknown to me, the ROOT prompt now seems to have
a different way of handling this case because just reverting commit
d97e4dca363 still works there...)
Instead of using relying on recursive templated calls, perform the
type cast in the ValueExtractionSynthesizer. This has the advantage
of avoiding an ODR violation warning in MultipleInterpreters.C with
LLVM 16 (while unclear if that one is correct or not).
Warning appears when building on Windows:
```
Diagnostics.cpp
C:\git\root\interpreter\cling\lib\Utils\Diagnostics.cpp(37,27): warning
C4805: '|': unsafe mix of type 'bool' and type
'int' in operation
[C:\Soft\root_64\interpreter\cling\lib\Utils\obj.clingUtils.vcxproj]
```
These symbols may not be found automatically. See also upstream issue
https://github.com/llvm/llvm-project/issues/61289. This fixes the
test DynamicLibraryManager/cached_realpath.C, approach by Lang Hames.
This new release includes some improvements:
* Add experimental support for forward vector mode
* Add pushforwards for std::floor and std::ceil
* Improve AD function interfaces with bitmasked options. For example:
clad::differentiate<<clad::order::first, clad::opts::vector_mode>(f) will
be equivalent to clad::differentiate<<1, clad::opts::vector_mode>(f) and
will request the first order derivative of f in forward vector mode.
* LLVM16 Support
See more at: https://github.com/vgvassilev/clad/blob/v1.2/docs/internalDocs/ReleaseNotes.md
Fixes https://github.com/root-project/root/issues/9449
so it is possible to define operator overload in cling cmd prompt.
btw, Make SkipPointerRefs() to not assume next token being a
tok::raw_identifier since it can be a global-scoped identifier,
e.g. int* ::class_a::func_b(short c) { return nullptr; }
