GitLab

Have QTextStream destructor care for the flushing also, unless needed

GIT_SILENT

GIT_SILENT

GIT_SILENT

Any linebreaks per message for output on console are added by the
logging/message handlers

GIT_SILENT

Fixup to 2b9608e1

GIT_SILENT

GIT_SILENT

Simplify the implementation of ClangUtils::getRawContents() by using clang_getFileContents(), increase the libclang minimum version requirement accordingly, and remove code for older versions of libclang

Signed-off-by: Juraj Oravec <sgd.orava@gmail.com>

This allows the toolview to take less horizontal space.
Signed-off-by: Juraj Oravec <sgd.orava@gmail.com>

Signed-off-by: Juraj Oravec <sgd.orava@gmail.com>

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

BUG:414978
FIXED-IN:5.4.6

GIT_SILENT

(cherry picked from commit faded356)

GIT_SILENT

GIT_SILENT

GIT_SILENT

GIT_SILENT

GIT_SILENT

GIT_SILENT

This removes the duplicate items from the code completion list
and unbreaks TestCodeCompletion::testOverrideExecute when running
against clang 9.

Finally no more failing tests in the clang plugin :)
Still ton's of things to do though :(

Seems like clang lost its smartness when completing switch labels.
It suggest non-constant expressions suddenly, which isn't even
valid C++. We need to fix that upstream somehow.

We only want to move definitions, not declarations - the code
checked the wrong thing. Additionally, it turns out that we actually
didn't set the 'isDefinition' correctly on inline class functions.

Fixes test_assistants

The upstream bug isn't resolved, and its behavior is still
broken, yet the test case works properly for us now. Potentially
this is related to some of the other options we are setting.
KeepGoing alone doesn't seem to do the trick though.

We now look at the return type of functions we complete, instead of
just adding the functions themselves. This sounds like a valid
approach but in the end isn't sufficient to solve all cases - we
really need to expand the libclang API to get this done properly.

Most notably, since self-assignment isn't allowed anymore, we cannot
find the correct type for the `int* p = ...` case anymore and just
fail then... For now, we will mark the test as an expected failure.

Recent versions of clang won't suggest self-assignment anymore.
I.e. previously when one wrote

    foo bar =

one would get `bar` as a code completion suggestion, which is
obviously broken. Adapt the tests now that this is correctly
filtered out within clang.

This may break testing on older libclang, I don't care.

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"