- 04 Dec, 2021 1 commit
-
-
Bernd Buschinski authored- on my system meson is a link to python2-exec $ ls -ld $(which meson) /usr/bin/meson -> ../lib/python-exec/python-exec2 The python2-exec must not be executed directly. - Currently the mesonExecutable is canonicalized, which also resolves symlinks This causes the meson plugin to execute stuff like: $ python2-exec --reconfigure instead of $ meson --reconfigure Fix this by not canonicalizing the mesonExecutable, validity is already checked earlier. BUG: 412477 FIXED-IN: 5.8.220400
-
- 03 Dec, 2021 3 commits
-
-
Igor Kushnir authoredhttps://ericniebler.com/2013/08/07/universal-references-and-the-copy-constructo describes the issue solved by this commit. The blog post recommends constraining the forwarding-reference constructor. But that solution would require much more code than added in this commit. Especially because the constraint would have to be expanded to support template parameter pack. The deleted constructor overloads prevent surprising and undesirable compilation of the following two code snippets: ScopedDialog<QDialog> s; ScopedDialog<QDialog> s2(s); and const ScopedDialog<QDialog> s; ScopedDialog<QDialog> s2(std::move(s)); They were compiled successfully because the forwarding-reference constructor was a better match than the deleted copy and move constructor respectively (due to const-qualifier differences). In the forwarding-reference constructor invoked in these snippets, a reference to ScopedDialog `s` was implicitly converted to QDialog* and passed as a parent to `s2`'s new QDialog. The final specifier prevents surprising and undesirable compilation of the following code snippet (as well as the 3 other const and rvalue-reference variations of it): ``` template <typename T> class ScopedDerived : public ScopedDialog<T> {}; void f() { ScopedDerived<QDialog> s; ScopedDialog<QDialog> s2(s); } ``` A volatile ScopedDialog (which does not make sense, by the way) cannot be copied or moved for another reason: passing volatile ScopedDialog as `this` argument to `operator DialogType*() const` discards qualifiers.
-
Igor Kushnir authored
Move constructor and move assignment operator have to be defaulted or deleted to follow the rule of five. ScopedDialog is supposed to live inside a specific scope, so moving it doesn't make sense. A similar class template QScopedPointer is also not movable. So let us define the move operations as deleted via Q_DISABLE_COPY_MOVE. Unlike Q_DISABLE_COPY, Q_DISABLE_COPY_MOVE prevents the forwarding constructor from acting as a move constructor. So the following code ScopedDialog<QDialog> s; ScopedDialog<QDialog> s2(std::move(s)); compiled before, but does not compile at this commit. Move this declaration closer to other special member functions. -
Igor Kushnir authored
This change improves performance by doing less copying during construction. In addition the forwarding allows to pass non-copyable objects by reference.
-
- 27 Nov, 2021 1 commit
-
-
Igor Kushnir authored
GIT_SILENT
-
- 26 Nov, 2021 2 commits
-
-
The declaration of QMutex(Recursive) is deprecated. This fixes several deprecation warnings. Many declarations had to be patched in order to return the correct constructs, because QRecursiveMutex is not convertible to QMutex.
-
The loop in DocumentParsePlan::removeTargetsForListener() caches m_targets.cend(), which can become invalidated when a target is erased inside this loop. Comparing to the invalidated end iterator leads to undefined behavior. This undefined behavior makes QmlJS's test_files crash every other time, both on my system and on the CI. The alternating crashes of this test can be seen by clicking Next Build repeatedly starting from the first build since the bug was introduced: https://build.kde.org/job/KDevelop/job/kdevelop/job/kf5-qt5%20SUSEQt5.15/165/testReport/junit/projectroot.plugins.qmljs/tests/test_files/ KDevelop often crashes because of this undefined behavior too (see the bug report referenced below). m_targets is a QSet, which is implemented in terms of QHash. QHash's end iterator `e` equals the d-pointer `d` via anonymous union. So the only way cend() can be invalidated is if QSet::erase() detaches. That's possible, because an entire DocumentParsePlan is copied in BackgroundParserPrivate::parseDocumentsInternal(): const DocumentParsePlan parsePlan = *parsePlanConstIt; The involvement of multithreading explains why test_files crashes every other time rather than each time. Examining call stacks of both test_files and kdevelop crashes confirms this hypothesis: a background thread crashes in DocumentParsePlan::removeTargetsForListener() while the main thread waits on the `m_mutex.lock()` line in parseDocumentsInternal(). This bug was introduced in 5ee9b9fe. BUG: 445699
-
- 24 Nov, 2021 1 commit
-
-
Script Kiddy authored
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"
-
- 23 Nov, 2021 1 commit
-
-
Ben Cooksley authored
-
- 21 Nov, 2021 1 commit
-
-
Ben Cooksley authored
-
- 17 Nov, 2021 1 commit
-
-
Igor Kushnir authored
-
- 16 Nov, 2021 1 commit
-
-
Eugene Shalygin authoredDo not repeat shell test in the kdevplatform_shell_environment.sh file, but instead detect the shell which runs the script via ps and $$. The testing and shell selection is done by the app/kdevelop! script. Fix tests for executables: which is a builtin in zsh and output redirestion does not work with it. Replace such tests with command -v, for which POSIX defines behaviour and return values. BUG: 442481 FIXED-IN: 5.7.0
-
- 15 Nov, 2021 3 commits
-
-
Previously, the goto source shortcut for summary diffs (e.g. diffs showing all staged changes) would not work. The reason was that the view data structure member url contained only the path to the project root. This commit changes 1. modifies the vcsDiff api diffLineToSource/Target to return relative source paths together with source lines 2. uses the result together with the project root to compute the url to the path in the shortcut handler. -
Script Kiddy authored
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"
-
Script Kiddy authored
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"
-
- 13 Nov, 2021 5 commits
-
-
Igor Kushnir authored
Initialize m_plugin with nullptr. Simplify the expression assigned to perforceTestBaseDir. The temporary directories are created before and removed after each test function. This makes switching to QTemporaryDir more difficult and less useful => retain QDir::tempPath() for the time being.
-
Igor Kushnir authored
These directories were created and removed but never used. PerforcePluginTest::removeTempDirsIfAny() attempted to remove perforceTestBaseDir twice because of what was almost certainly a typo. perforceTestBaseDir2 was supposed to be removed there, but it is better to never create and never remove this unused directory.
-
Igor Kushnir authored
ItemRepositoryRegistry::deleteRepositoryFromDisk() only marks the repository for removal. Call ItemRepositoryRegistry::shutdown() to actually remove the files when done testing. Even though QTemporaryDir already removes the directory with all repository files in it, let us shut down the repositories properly to test their cleanup too.
-
Igor Kushnir authored
The temporary directory and files are now removed at test/bench exit.
-
Igor Kushnir authored
test_qmakeproject creates /tmp/some/path. After the test exits, this directory still exists, contains an empty directory "build" and two non-empty files (.qmake.stash and Makefile). This change brings two improvements: * the temporary directory path is no longer hardcoded; * the temporary files and directories are removed at test exit.
-
- 12 Nov, 2021 10 commits
-
-
Igor Kushnir authored
The performance gains are not very significant, but the code becomes shorter as well, so there seem to be no drawbacks to this change. Average BenchLanguageController results before and at this commit in milliseconds per iteration: Data row Before At 1. benchLanguagesForUrlNoCache() CMakeLists 0.00045 0.00046 cmakelists wrong case 0.00045 0.00046 lower-case 0.00045 0.00040 upper-case 0.00045 0.00039 mixed-case 0.00045 0.00040 .C 0.00037 0.00040 .cl 0.00038 0.00038 existent C with extension 0.00043 0.00045 .cc 0.00040 0.00039 .cmake 0.00041 0.00044 .diff 0.00038 0.00040 .qml 0.00037 0.00039 existent C w/o extension 0.16 0.16 existent patch w/o extension 0.20 0.20 2. benchLanguagesForUrlFilledCache() CMakeLists 0.00048 0.00050 cmakelists wrong case 0.00049 0.00050 lower-case 0.00048 0.00041 upper-case 0.00049 0.00041 mixed-case 0.00049 0.00042 .C 0.00041 0.00042 .cl 0.00045 0.00042 existent C with extension 0.00047 0.00047 .cc 0.00044 0.00041 .cmake 0.00042 0.00043 .diff 0.00041 0.00041 .qml 0.00045 0.00043 existent C w/o extension 0.16 0.16 existent patch w/o extension 0.20 0.20 Note: benchLanguagesForUrlNoMatch*() benchmark results are practically unaffected by this commit because the QMimeDatabase::mimeTypeForFile() call is orders of magnitude slower than the MimeTypeCache::languagesForFileName() call optimized here. The benchmarks run slightly faster on average now. benchLanguagesForUrlFilledCache(), which is closest to the real-life LanguageController usage in KDevelop, got a 14% speedup for the common *.cpp pattern data rows ("lower-case", "upper-case" and "mixed-case"). -
Igor Kushnir authored
Previously the filename of each url argument to LanguageController::languagesForUrl() was matched against all cached glob pattern suffixes (sometimes not "all" as suffixes of an already matched/added language were skipped). Now the filename is only matched against the suffixes that end with the same case-insensitive character. This change should improve the performance of this frequently called function, because case-insensitive string comparison looks like the most expensive among the actually performed in practice operations in cases when the language is found in mimeTypeCache. Suffix matching performance could be further improved by storing reverse suffixes in a trie (aka prefix tree). However I could not find a suitable trie implementation in KDevelop or one of its dependencies. QmlJS::PersistentTrie seems to be the closest match, but it has many unneeded features and stores only strings with no ability to attach user data to them (such as ILanguageSupport*). Implementing a trie data structure specifically for this optimization would be an overkill. For some reason, using std::lower_bound in place of std::equal_range in MimeTypeCache::languagesForFileName() makes the most important benchmark - benchLanguagesForUrlFilledCache() - somewhat slower on most data rows. I have implemented and benchmarked 3 alternative optimizations of suffix matching in MimeTypeCache::languagesForFileName(): 1. Binary searching for the last fileName's extension in the std::vector<StringLanguagePair> sorted by the entire last extension string case-insensitively. 2. Binary searching for lower-cased last fileName's extension in the lower-cased std::vector<StringLanguagePair> sorted by the entire last extension string case-sensitively. 3. Looking up lower-cased last fileName's extension in lower-cased QMultiHash<QString, ILanguageSupport*>. The second alternative optimization turned out to be faster than the first one; the third - faster than the second one. All 3 alternative optimizations turned out to be: a) Less general than this version: multiple extensions, such as tar.gz, were no longer optimized. This is only a theoretical downside because none of the maintained KDevelop plugins supports wildcard patterns with multiple extensions. b) Slower than this version. Probably because of expensive temporary QString construction - the (lower-cased) last fileName's extension. There is currently no need to optimize literal pattern matching, because only one literal pattern is supported: "CMakeLists.txt". If more literal patterns become supported in the future, the elements of MimeTypeCache::m_literalPatterns can be easily lower-cased and sorted to speed up matching against them. Average BenchLanguageController results before and at this commit in milliseconds per iteration: Data row Before At 1. benchLanguagesForUrlNoCache() CMakeLists 0.00046 0.00045 cmakelists wrong case 0.00046 0.00045 lower-case 0.00058 0.00045 upper-case 0.00058 0.00045 mixed-case 0.00058 0.00046 .C 0.00050 0.00039 .cl 0.00070 0.00038 existent C with extension 0.00053 0.00044 .cc 0.00058 0.00041 .cmake 0.00039 0.00041 .diff 0.00037 0.00038 .qml 0.00036 0.00037 existent C w/o extension 0.16 0.16 existent patch w/o extension 0.20 0.20 2. benchLanguagesForUrlFilledCache() CMakeLists 0.0011 0.00050 cmakelists wrong case 0.0011 0.00050 lower-case 0.00083 0.00048 upper-case 0.00083 0.00048 mixed-case 0.00085 0.00048 .C 0.00072 0.00040 .cl 0.00091 0.00042 existent C with extension 0.00064 0.00046 .cc 0.00080 0.00043 .cmake 0.00090 0.00042 .diff 0.00083 0.00041 .qml 0.00093 0.00045 existent C w/o extension 0.16 0.16 existent patch w/o extension 0.20 0.20 3. benchLanguagesForUrlNoMatchNoCache() empty 0.0016 0.000013 4. benchLanguagesForUrlNoMatchFilledCache() empty 0.0018 0.000013 Note: benchLanguagesForUrlNoMatch*() benchmark results are practically unaffected by this commit because the QMimeDatabase::mimeTypeForFile() call is orders of magnitude slower than the MimeTypeCache::languagesForFileName() call optimized here. The only exception is the "empty" data row, which became more than 100 times faster thanks to the added early return. Most benchmarks run faster now. benchLanguagesForUrlFilledCache(), which is closest to the real-life LanguageController usage in KDevelop, runs about two times faster thanks to this commit. -
Igor Kushnir authored
Bugs in the old implementation that are fixed in this commit: * The regular expression matching code was incorrect and so it never matched the only pattern it handled: "CMakeLists.txt". As a result, LanguageController::languagesForUrl(".../CMakeLists.txt") returned an empty language list in background threads; resorted to extra work culminating in a call to LanguageController::languagesForMimetype() in the main thread. * The suffix matching optimization matched patterns case-sensitively and so didn't match names like "X.CPP". As with the regular expression matching bug, this resulted in a wrong return value in background threads and extra work in the main thread. * The suffix matching optimization assumed that '*' is the only wildcard character. While this is actually the case for glob patterns of the mime types that can currently end up in mimeTypeCache, it might have led to a surprising bug if more complex glob patterns became supported in the future. TestLanguageController::testLanguagesForUrlWithCache() fails on the following data rows without this commit because of these bugs: - CMakeLists - cmakelists wrong case - upper-case - mixed-case Improvements of the MimeTypeCache reimplementation in this commit: * Literal pattern optimization: "CMakeLists.txt" was the only pattern that required regular expression matching in the old implementation. It could not be handled by the suffix matching optimization. Now the new separate pattern category m_literalPatterns handles this case so that the slower regular expression matching never happens in practice. * The suffixes, literal patterns and regular expressions are now created once and cached rather than constructed in each languagesForUrl() call. * QRegularExpression is now used instead of the deprecated QRegExp. This is the list of wildcard patterns supported by maintained KDevelop plugins (collated from X-KDevelop-SupportedMimeTypes plugin entries and /usr/share/mime/globs2): kdevclangsupport text/x-chdr *.h text/x-c++hdr *.hh *.hpp *.hp *.h++ *.hxx text/x-csrc *.c:cs text/x-c++src *.c++ *.cc *.cxx *.C:cs *.cpp text/x-opencl-src *.cl text/vnd.nvidia.cuda.csrc *.cu text/vnd.nvidia.cuda.chdr *.cuh text/x-objcsrc *.m kdevpatchreview text/x-patch *.patch *.diff kdevqmljs text/x-qml *.qml *.qmlproject *.qmltypes application/javascript *.jsm *.mjs *.js KDevCMakeManager text/x-cmake cmakelists.txt *.cmake KDevCssSupport text/css *.css text/html *.html *.htm KDevPhpSupport application/x-php *.phps *.php *.php3 *.php4 *.php5 kdevpythonsupport text/x-python *.wsgi *.py *.pyx text/x-python3 *.py3x *.py3 *.py KDevRubySupport application/x-ruby *.rb Only *.c and *.C out of all supported patterns should be matched case-sensitively. But both of these patterns belong to the same plugin - kdevclangsupport. So LanguageController can safely match all patterns case-insensitively. See also https://specifications.freedesktop.org/shared-mime-info-spec/shared-mime-info-spec-latest.html Average BenchLanguageController results before and at this commit in milliseconds per iteration: Data row Before At 1. benchLanguagesForUrlNoCache() CMakeLists 0.029 0.00046 cmakelists wrong case 0.029 0.00046 lower-case 0.0023 0.00058 upper-case 0.029 0.00058 mixed-case 0.029 0.00058 .C 0.0023 0.00050 .cl 0.0023 0.00070 existent C with extension 0.0022 0.00053 .cc 0.0023 0.00058 .cmake 0.0016 0.00039 .diff 0.00094 0.00037 .qml 0.0012 0.00036 existent C w/o extension 0.16 0.16 existent patch w/o extension 0.20 0.20 2. benchLanguagesForUrlFilledCache() CMakeLists 0.032 0.0011 cmakelists wrong case 0.031 0.0011 lower-case 0.0039 0.00083 upper-case 0.030 0.00083 mixed-case 0.030 0.00085 .C 0.0039 0.00072 .cl 0.0039 0.00091 existent C with extension 0.0038 0.00064 .cc 0.0039 0.00080 .cmake 0.0039 0.00090 .diff 0.0039 0.00083 .qml 0.0039 0.00093 existent C w/o extension 0.16 0.16 existent patch w/o extension 0.20 0.20 3. benchLanguagesForUrlNoMatchNoCache() empty 0.0021 0.0016 archive 0.024 0.023 OpenDocument Text 0.024 0.023 existent archive with extension 0.030 0.029 existent archive w/o extension 0.15 0.15 4. benchLanguagesForUrlNoMatchFilledCache() empty 0.0054 0.0018 archive 0.029 0.024 OpenDocument Text 0.029 0.024 existent archive with extension 0.035 0.030 existent archive w/o extension 0.16 0.15 Almost every benchmark runs faster now. Many run more than ten times faster thanks to this commit. -
Igor Kushnir authored
LanguageController's initialize() and cleanup() member functions were called once each by Core and didn't reset LanguageController's state. This commit expands the definition of initialize() to make these two member functions work for the added tests. Most of the added code belongs in cleanup(), but moving it there seems a bit risky: what if this data is used at KDevelop exit?
-
Igor Kushnir authored
GIT_SILENT
-
Albert Astals Cid authored -
Albert Astals Cid authored
-
Before, when an element in one of the categories (staged/unstaged/...) was selected, its context menu didn't have the refresh repo action. This was not very convenient, since one had to first unselect the element and then trigger the context menu again, when the repo model was out of date. This commit fixes this minor annoyance. (cherry picked from commit d660f70c)
-
Script Kiddy authored
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"
-
Script Kiddy authored
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"
-
- 11 Nov, 2021 7 commits
-
-
Igor Kushnir authored
-
Igor Kushnir authored
ILanguageSupport::indentationSample() is used in a single place: CustomScriptPlugin::computeIndentationFromSample(), which in turn is used only in CustomScriptPlugin::indentation(). When indentationSample() returns an empty string (as the base class implementation does), CustomScriptPlugin::indentation() fails and returns invalid Indentation. f31d32f4 removed CppLanguageSupport, which was the only class that overrode indentationSample(). Not copying this override into ClangSupport was probably a mistake. This commit copies the last version of CppLanguageSupport::indentationSample() into ClangSupport::indentationSample(). The copied sample matches the example in the documentation for ILanguageSupport::indentationSample().
-
Jonathan L. Verner authored
-
Jonathan L. Verner authored
Upon startup, when no project was initially selected, the commit form was previously enabled (however, pressing the commit button didn't work). This commit initially disables the commit form and only enables it after a user has selected a project from the tree in the git commit toolview.
-
Jonathan L. Verner authored
When committing fails, the user should be informed (and not expected to notice this by seeing an error in the Version Control output view or knowing commit message not being cleared means failure). I opted to use a KMessageWidget instead of a Modal Dialog (per the HIG guidelines, although they are a bit unclear, since there is no clear action that the user can take to remedy the situation). I also opted for just a short message pointing the user to the Version Control output view for more information due to the limited space available in the commit toolview. BUG: 443475 FIXED-IN: 5.7.0
-
-
Jonathan L. Verner authored
This fixes a bug which would lead to the commit form being disabled and unusable when a commit failed for some reason. The commit form was disabled when the commit job started and was meant to be enabled when the job finished. However, the wrong signal was connected to, which meant that in case of failure the code never ran. BUG: 443475 FIXED-IN: 5.7.0
-
- 10 Nov, 2021 3 commits
-
-
Before, when an element in one of the categories (staged/unstaged/...) was selected, its context menu didn't have the refresh repo action. This was not very convenient, since one had to first unselect the element and then trigger the context menu again, when the repo model was out of date. This commit fixes this minor annoyance.
-
Igor Kushnir authored
This should be faster than copying SourceFormatterStyle at each call of this function. Introduce a QString styleContent variable in CustomScriptPlugin::formatSourceWithStyle() for these reasons: * style is now a reference to const and cannot be assigned to; * the name of an incomplete style matches the name of the corresponding predefined style, so using unmodified style.name() below is correct; * QString content() is the only other member of style used below. Reformat the lines touched by this change. Don't refer to the previous name of kdevcustomscript plugin "indent" in the warning message touched by this change.
-
I'm working with a project that has more than 1400 Unit-Tests and they are all in a single executable. This causes CTestFindJob to queue each c++ File compiled into the test executable more than 1400 times. We might probably want to change something about that, but that is a bit more work; Therefore I chose to get the low hanging fruits. These changes improve the queuing performance by a factor of approx. 10. For my project this means roughly 3 Seconds of waiting instead of more than 30 Seconds before the change.
-
