The string cache is only very seldom written but very often read. To
improve thread scaling it is faster to lock it only for write operations.
So we use a shared mutex which is locked in shared mode for read
operations and locked exclusively for write operations.
Change-Id: I2cc742d1b9cc15c162be40ab339fa8310640bc44
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
The database is using file path integer ids to handle file paths because
otherwise we would save many redundant data. This patch is improving it
further with the introduction of a database based file path cache. The
entries are now divided in a directory path and file name. This is quite
handy for directory based file watching.
Change-Id: I03f2e388e43f3d521d6bf8e39dfb95eb2309dc73
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Now it's in accordance with clangrefactoringbackend/source and
clangpchmanagerbackend/source.
Change-Id: I939cfc72cffb8fcde0649f125e6efb6670d1cbc3
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
llvm-config-3.9 --cxxflags has -g1 which is substituted by 1.
Change-Id: I626f3f15c659f7de030c08883b5b739b1d636ada
Reviewed-by: Marco Bubke <marco.bubke@qt.io>
Make build command the same for all builders.
Minimize differences.
Change-Id: I1cfe5071b3afb4944ed178fff1e57d3aee45d8a9
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
Make this highlighting work without builtin code model
but based on the HighlightingMarks that we already have
from ClangCodeModel.
Redundant parameters are removed by this change.
Change-Id: I73b5dab46ba59d2f813236831818f0a9bc94c5bc
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
Do what can be done in the constructor instead of relying on the
clients.
Change-Id: I5f475b5309afb37e34d228e54ad3c12f1698d72b
Reviewed-by: Ivan Donchevskii <ivan.donchevskii@qt.io>
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
There are now only value and values methods. value returns an optional
and values are returning a vector. The result value count has now be
specified instead of the result value list.
Change-Id: I17a0741d5e838b4bf4b9486825c870ada1722584
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
It is required by many template function because it is used in the STL.
Change-Id: I5a2b9266d0c9d79bdc37cbbc4e7b62fac7a82b8b
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
lastInsertedRowId() function was added and more test are now under test.
Change-Id: I02bf11dbab29654dbff9f2cad8c13c0c4d15e3be
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
We moved the creation of the tables and indices to an extra class which can
be called from other places. So you can be sure that a database is
initialized.
Change-Id: Ief5b30ced7b9011ca94367aa2578098423dcecd9
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
If you use the database in a multi-threaded environment, you must always
your statements in transactions.
For read-only statements, you use DeferredTransaction and write statements
you use ImmediateTransaction. If you mix read and write statements you have
to use ImmediateTransaction. Don't use DeferredTransaction because it leads
to undefined behavior.
Change-Id: Ida298a20f33423c8da09e768a7b658dd8e918f46
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Reviewed-by: Marco Bubke <marco.bubke@qt.io>
We now have the very powerful write and read methods, so using the low
level interface leads only to stupid errors and prevents optimizations.
Change-Id: If2340833977935cb2d319735dee96b0bac33142b
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Reviewed-by: Marco Bubke <marco.bubke@qt.io>
Introducing different exceptions for different error cases.
Change-Id: I4371d1e64d9dca2a9f68dcbaa4a891c55879c1f5
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Reviewed-by: Marco Bubke <marco.bubke@qt.io>
We use a name space now, so the prefix is not of much use.
Change-Id: I2b077576f94dab778add6ab2e54870f7ca18da78
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
The reserve function is quite complicated because it is implementing a
grow pattern. Something we seldom need. So we now align to the next cache
line size.
Change-Id: I14bb88c12bd740a7afa7cd08969a4e07fb0f9add
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Under windows the signature for private and public member are different. So
the hack to make private members public is not working. The other method
to use a macro project macro header is not wanted.
Change-Id: I25a3f4a8645394ba6b49a7d4778adab90155f892
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Create an interface to get the ability to use
another FollowSymbol implementation
Change-Id: I5802f62523ff3ee47b8a14e487adf43edcb6c9b1
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
..and use it when we don't have refactoring plug-in
Change-Id: Ibe317a9728d439b9c5e05271d92a330d22eaacb9
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
The old code model expected the macros as C++ formatted text
("#define Foo 42) but newer targets like the Clang codemodel expect key
value arguments like "-DFoo=42". So instead of parsing the text again and
again we use an abstract data description.
Task-number: QTCREATORBUG-17915
Change-Id: I0179fd13c48a581e91ee79bba9d42d501c26f19f
Reviewed-by: Tobias Hunger <tobias.hunger@qt.io>
It is a first step and now a database is generated if you start QtCreator.
Some code is now shared with the PchManager which can be improved in the
future.
Change-Id: Ic267fe7960f6c455d91832859a673ce98f269aa2
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
If a parse or reparse operation fails, the corresponding translation
unit is "not intact" anymore and further job requests for that
translation unit will be silently discarded without notifying the
plugin. This is especially problematic when the plugin expects/waits
(for) a certain message.
A parse/reparse can fail due to e.g. an invalid commandline or an
internal crash.
In such a case, ensure to send some dummy results.
Task-number: QTCREATORBUG-18864
Change-Id: Ida9b8066fd55e64027a9b0d8bd5cf365b16a7356
Reviewed-by: Ivan Donchevskii <ivan.donchevskii@qt.io>
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
Change unsaved file to provide constructor overloads
for std::make_unique, std::make_shared and
QSharedPointer::create
Example:
// Provide Foo constructor signatures at <Cursor>
std::make_unique<Foo>(<Cursor>
Task-number: QTCREATORBUG-18615
Change-Id: I87dd17085adf99ee498db969a3cdda5ebd973476
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
... to reuse this position in backend (instead of
searching the function start again)
Change-Id: I02818dce4fc37ed6e7ecfb533191dbfe60610204
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
The old algorithm was comparing the string two times. One timer for smaller
and one for larger. And memcmp on a long string still costs. The new one
has a three state so it can compare smaller, greater and equal in one.
There is a reverse version too which has big advantage for file paths.
Change-Id: Ica4024f0a071803c697e2c1f26edd3eb1b203f9f
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
We already share same classes there which has nothing to do with IPC and
I want to more for sharing. So we should use a name which fits better.
Change-Id: Idfb12b6de714206117b92634ad719c6a0e290e78
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
It is now possible to read values at once.
for (auto [name, value] : statement.tupleValues<String, int>(1000, "foo", 20))
....
Change-Id: I3d4bc5218810b4620e1df625126aa490f30bbc71
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
There is now a forward header and it is possible template the string type.
Change-Id: Ibebd32c475a2d89a90da7190e77d44278eb43740
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
It is more efficient to use compare instead of less and equal together.
Change-Id: I772b03ba02509c0ebc5d5d770be14cf7d5bd431e
Reviewed-by: Tim Jenssen <tim.jenssen@qt.io>
The server and client interfaces was tightly coupled.So it prevented
the introduction of immediate interfaces.
Change-Id: Ie4197ffddb862e076d080b3d2a5ee869fad9e4d0
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@qt.io>
Currently this might happen when registerTranslationUnitsForEditor is
called with visible documents that are not the ones that are registered.
Change-Id: I9ae5f75c8bbff6a11161a3387633726066001062
Reviewed-by: Marco Bubke <marco.bubke@qt.io>
