Advapi32Util.registryGetValues() is more efficient to fetch all the
values since the regiter only needs to be opened once.
XXX: Due to a cycle in the dependencies we need to split this
in two parts, this previous commit handles the core.native API change
and this commit handles the use of the new API
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
Change-Id: I960954e332e05dfebcd0263f983bbbc6c64d46af
Advapi32Util.registryGetValues() is more efficient to fetch all the
values since the regiter only needs to be opened once.
XXX: Due to a cycle in the dependencies we need to split this
in two parts, this commit handles the core.native API change
and the next commit handles the use of the new API
Change-Id: Ifd9f1ccc44c652ef3b517278bd342a486155c5fb
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
Use the WindowsRegistry implementation rather than having yet another
implementation for reading the Windows registry in native code.
Change-Id: If12068319ea3b99934112208a0a21538c792909c
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
When unchecking the "Use default build command" checkbox, hide the GUI
options to configure the flags as the user should have full control.
This solution was discussed in
https://git.eclipse.org/r/c/cdt/org.eclipse.cdt/+/171735
Change-Id: Ifbd67b279db55f5a83f4a660b9a1248e5c021df2
Also-by: cartu38 opendev <cartu38.opendev@gmail.com>
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
ICPPUnaryTypeTransformation was meant to be used only when the
transformation is applied on a dependent type. But it was actually
always used when creating types for decl specifiers, regardless if a
dependent type was involved or not. The untransformed type was causing
issues because code dealing with ITypes doesn't apply the transformation
everywhere. It seems better to apply the transformation early when
possible and let the rest of the logic intact.
Change-Id: I1b6d77a857e901f71f00e935e75d24cea87c3118
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
Don't define __STDC__ for clang-cl, just like msvc. Added a new compiler
type to differentiate clang-cl from clang in order to be able to add the
define just for clang and not clang-cl. I initially thought that the
unresolved symbol was caused by a missing built-in symbol (_alloca) so I
started to implement MSVC as a separate language from GNU and not
pollute GCCBuiltinSymbolProvider with a new concept of compiler type,
etc. But it turns out _alloca is also in declared in the headers and was
inactive because of the wrongly defined __STDC__. So this version of the
fix of adding a compiler type to GPPScannerExtensionConfiguration is
much less intrusive and risky, although adding a new language could have
its merit in case of more substantial change.
Change-Id: Ieb27b7dbcc531b8fd3fe30777cd7f343fd3ba66f
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
IType.isSameType was doing pretty much already what was needed. Added
GCC 6.0 and 10.0 parser configs to enable these built-ins for the proper
versions.
Change-Id: Ifd2908e726c098fb07c9420b29e2cb26014419bf
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
All compile warnings for native code should be handled.
Change-Id: I5ee391fd24339cb55646dbd2980d2856fabe8ddb
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
After loading the spawner library, call the native method
configureNativeTrace() that will query the Eclipse platform for several
debug option strings.
Change-Id: I031bb2cdc04ba2675913b3b2e320039c28139638
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
In cases where the dwarf headers say there is data at a position that
does not exist, the position call throws an exception. Prior to this
patch that exception, being a runtime exception, can cause the whole
IDE to exit in the case that the standalone debugger is used.
Change-Id: I7e12667890490a49ecd63785ea2cc7c02c08679d
* Unify pointer checkes
* Avoid using negated conditions.
* Reduce scope of local variables when possible
Change-Id: Ibacd13126351019af544f3e22513654d5ffee342
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
This will enforce formatting when building the native code
Change-Id: I6c047f4c0672609df322b7ba716fc786f0e3aab4
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
The comment is part of the generated header file
Change-Id: Ie890ad4d906c0f4e6a23b2a42a1ef342d1da8865
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
Generalize the type of elements and resolve type safety warnings
Change-Id: Id8e443d0c071bbd5447092099e80574cdf7fd9b3
Signed-off-by: Alexander Fedorov <alexander.fedorov@arsysop.ru>
Add /clang:-isystem. It's handled similarly to /imsvc but interestingly it
doesn't support having spaces between the option name and it's value.
Change-Id: Ic5d61df1c7adebbf707d93a60858e942354c9ed9
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
Added a new C/C++ formatter profile called "Unmanaged profile 'CDT'"
(name stolen from the Java formatter) that is basically K&R with the
tweak of maxium line width set to 120 (same width as for .java files).
Added enforcement of the formatter during build.
libspawner.so/jnilib have differences because the re-formatting changes
line numbers and therefore the __LINE__ macro expands to a different
value.
Change-Id: Id3a0619cb31640c7817dc684c72139f90cab0fc6
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
List as many as possible of the failed bundles when doing the baseline
compare.
Change-Id: I0d2817c418cf865f7fc6ea1c7e25d5b3b7b41bd5
Signed-off-by: Torbjörn Svensson <azoff@svenskalinuxforeningen.se>
Generalize the type of elements and resolve type safety warnings
Change-Id: Ibbbdc26663a7fede9debef3ca5d1c79ca346608c
Signed-off-by: Alexander Fedorov <alexander.fedorov@arsysop.ru>
Generalize the type of elements and resolve type safety warnings
Change-Id: I5cb8d06a80207a7df8240f499936cdc49d49964c
Signed-off-by: Alexander Fedorov <alexander.fedorov@arsysop.ru>
Change the default workspace indexer setting
Use the active configuration by default, which is less confusing because
the UI is properly reflected on active config change. Using a fixed
config can be seen as a more advanced setting for performance concerns.
A new preference is added, only used at default scope:
org.eclipse.cdt.core/cprojectdescription.configRelations
This can be used by products to customize the default according
to their needs (plugin_customization.ini).
This was done because this is a fairly impactful change for users.
Change-Id: I35888ffe5bc1814943f432f88a12094394924c88
Signed-off-by: Alex Freidin <freidin.alex@gmail.com>
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
I don't think "inheriting" by hand the background colors is supposed to
be done like this. Removing the lines setting the background to be the
same as the parent fixes the issue. I also verified all places in the UI
before/after the change.
Change-Id: I2eb4cc5afdd303d5d5613fc3d50d67d0c18c7028
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
Use List<IProject> instead of raw List
Use HashSet<IProject> instead of raw HashSet
Change-Id: I583daad225f828345d7109dba582ad5507431d65
Signed-off-by: Alexander Fedorov <alexander.fedorov@arsysop.ru>
Use Map<String, String> instead of raw Map
Use Entry<String, String> instead of raw Entry
Change-Id: I6e9ffc4dad34fcb8d2b0b1db04cfc901d9225f74
Signed-off-by: Alexander Fedorov <alexander.fedorov@arsysop.ru>
Attributes `url` and `version` can be omitted for features
Change-Id: I37ef5f52cd3a22dc9ccbce0905aa6642edf60a39
Signed-off-by: Alexander Fedorov <alexander.fedorov@arsysop.ru>
CoreModelUtil.findTranslationUnit only returns CElement in the populated
CModel of a project. This shouldn't change as a large majority of client
code need to see the model this way and not consider files that are
outside source folders. So for a file not under a source folder (and
therefore not in the CModel), we can just create a new translation unit
instance for it. This is actually how the editor deals with it too.
Change-Id: I8898822e94cac8562edcc0a726fdd8680119faca
Signed-off-by: Marc-Andre Laperle <malaperle@gmail.com>
