| Age | Commit message (Collapse) | Author |
|---|
|
So that #include <kconfig_version.h> still works after changing its
location in the previous commit.
NO_CHANGELOG
|
|
|
|
The eliminated duplication of the composite condition was error-prone
and less readable.
|
|
entryMap is ordered by the group name first. So there is no need to
iterate over the entire map to process entries whose group names start
with some prefix. Find the group name prefix's lower bound and iterate
over the proper subrange instead. This should be much faster, especially
if the subrange's size is much less than the entryMap's size.
Adjust isGroupOrSubGroupMatch() helper function to assert the extracted
startsWith() condition instead of rechecking it. Pass
KEntryMapConstIterator in place of the group name to this function in
order to simplify its callers' code. Reuse this helper function in
KConfigPrivate::copyGroup().
|
|
|
|
|
|
Usually multiple key entries belong to each group. Before this commit
every key entry's group name was converted from UTF-8 to UTF-16 and then
inserted into a set. With this commit every key entry's group name is
inserted as a std::string_view into a set, and only unique group names
are converted to UTF-16 in the end. This should improve performance.
|
|
This commit is an alternative to the earlier incorrect and reverted
b3dc879e8b108c26c929bfbe551bcdf77f140e94. That commit contained several
mistakes and an algorithmic complexity increase:
1) the added conditions were inverted: should have been
`hasNonDeletedEntries` (without `!`);
2) KConfigPrivate::groupList() passed group instead of key.mGroup to
hasNonDeletedEntries();
3) The complexity of hasNonDeletedEntries() is O(entryMap.size()). Calls
to this function were added into loops that iterated entryMap.size()
times. So the overall complexity of groupList() increased from linear
to quadratic.
This fix collects `mGroup`s of non-deleted key entries instead of
`mGroup`s of group entries. The number of key entries can be much
greater than the number of group entries, so this fix hurts performance.
But at least the algorithmic complexity of groupList() stays linear.
Future commits can optimize the loops and make them almost as fast or
even faster than before this fix.
The `!key.mKey.isNull() && !entryMapIt->bDeleted` checks added in this
commit are consistent with the check in
KConfigPrivate::hasNonDeletedEntries(). KConfig::hasGroupImpl() forwards
its argument to hasNonDeletedEntries() with the following comment:
// No need to look for the actual group entry anymore, or for subgroups:
// a group exists if it contains any non-deleted entry.
BUG: 384039
FIXED-IN: 5.90
|
|
|
|
|
|
Both GCC and Clang don't require a const integral value to be captured
explicitly to be used in a lambda, but of course one compiler thinks
otherwise...
Thanks to Ömer Fadıl Usta for the heads up.
GIT_SILENT
|
|
This was built with:
-DQT_MAJOR_VERSION=6 \
-DEXCLUDE_DEPRECATED_BEFORE_AND_AT=5.90.0 \
-DKF_DISABLE_DEPRECATED_BEFORE_AND_AT=0x055a00
Move the include(KDEInstallDirs) call before the first find_package(Qt*,
the former is what auto-detects the Qt version, and defaults to 5. This is
needed to be able to build against Qt5 by default.
All unit tests still pass.
|
|
NO_CHANGELOG
|
|
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"
|
|
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"
|
|
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"
|
|
It is redundant since the parent class, KConfig, has locationType().
Add some API docs for fileName(), even though one could use KConfig::name()
from the parent class, I think this would be confusing given a .desktop file
has a Name= entry, so fileName() is less ambiguous.
These seem to be leftovers from a previous refactor that happened a long time
ago.
|
|
NO_CHANGELOG
|
|
|
|
|
|
|
|
This will output a warning if an invalid value is requested. The goal is to avoid
implicit conversion which might result in a zero-int value.
|
|
By using an enum we have a central place to provide docs for the most
common actions/restrictions. Also consumers can pass in type-safe enum
values instead of potentially undocumented strings that might contain typos.
Also it is better documents is a value is supposed to be authorized using
KAuthorized::authorize or KAuthorized::authorizeAction, in the case of "shell_access"
this was mixed up from time to time.
Considering that we do not want the parameter for the methods to become too long
I have deliberately decided against using `enum class`. Also we don't have
and usecases for the binary operators in combination with the newly added enums.
Task: https://phabricator.kde.org/T11948
|
|
This will become especially useful when moving state data from the config file
to a dedicated state data file.
Task: https://phabricator.kde.org/T12549
|
|
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"
|
|
KConfigIniBackend relies on QLockFile which changes the content Uri and
this is bound to fail because we don't have permission. So, for these
Uris we use Android's Internal cache directory to save the .lock files.
|
|
GIT_SILENT
|
|
We were calling open(), then fopen(), then QFile::open(). This patch
removes the fopen() call that does not seem to do anything.
|
|
|
|
|
|
Now that KF requires Qt 5.15.2; this basically reverts commit 2e8742e64fc0
with some trivial changes.
NO_CHANGELOG
|
|
clang-tidy checks:
readability-isolate-declaration and readability-braces-around-statements
KF task: https://phabricator.kde.org/T14729
GIT_SILENT
|
|
The code was using a QSet (hash-based), use a std::set instead
for unicity, it gives us sorting for free.
We probably want to do the same in groupList(), but that's separate.
|
|
This is the case where we expected to see Key[$d] in the config file,
and it was somehow broken. When saving, the key was omitted, so when
reloading the kdeglobals key was present again.
Detected when trying to write a unittest for a different patch...
I had to reshuffle the unittest a bit because testThreads calls
testSimple which didn't expect that the "[AAA]" group would actually
be deleted now.
|
|
GIT_SILENT
|
|
|
|
We are only interested in whether the file doesn't exist when it fails
to open. This saves a stat on every successful config file parse.
|
|
At the moment we are taking a bit of a detour by converting to char*.
Not a bit deal but reads better and ends up being less calls.
|
|
Whenever we read a config file, we are first parsing kdeglobalsrc. This
caches the parsed entries so that they can be reused as long as it
hasn't changed.
This reduces the parseConfig calls (which is mostly prominent at
early process startup) from 18 to 12 (33%) ksmserver-logout-greeter.
This also means there's better cache locality as well as less memory
allocated as a lot of objects are shared but I have not measured this.
|
|
NO_CHANGELOG
|
|
This reverts commit b3dc879e8b108c26c929bfbe551bcdf77f140e94.
This change breaks plasmashell startup and possible other apps as well.
Reverting so the root cause can be investigated without time pressure.
|
|
calling deleteGroup only deletes all entries but the group does still
exists in listGroups
this is somewhat irritating since calling exists on that group will
return false
with this patch the group does no longer exists in listGroup
|
|
NO_CHANGELOG
|
|
QStringView has some bits of API that were only added in 5.15.2, whereas KF
requires 5.15.0.
This reverts commit 1780fb2a237af80ddc1f9cfb70cb892b53b91990.
|
|
The symbol export/visibility attribute is not standardized so far and
needs to be set by language extension attribute declaration at least
with clang & MSVC, who both support that when it appears before the
standard attribute declaration.
NO_CHANGELOG
|
|
NO_CHANGELOG
|
|
Some methods in ConfigLoaderHandler always returned true, change them to return
void instead. Also port them to take a QStringView instead of QStringRef, this
doesn't require a lot of changes because a QStringView can be constructed from
a QStringRef.
QXmlStreamAttribute methods like value() and name() return QStringRef in Qt5
and QStringView in Qt6, "fix" the issue by using auto keyword, which works
in both cases.
QStringView::toInt() isn't efficient in Qt5 so make the build conditional.
NO_CHANGELOG
|
|
This matches the method actually does, and makes it clearer that you don't
need to do config->sync() followed by config()->reparseConfiguration(), the
latter should suffice.
|
|
|
|
Since the FSDevice .desktop template hasn't been installed for a long time,
this method in KDesktopFile hasn't been useful. The whole FSDesktop concept
seems to be redundant nowadays, since CDRom and Floppy drives have been replaced
by USB sticks.
See this KF6 taks for more details:
https://phabricator.kde.org/T14295#253759
|
