| Age | Commit message (Collapse) | Author |
|---|
|
Dolphin shows the size of directories by listing their contents, which
for some users might cause unwanted load on the file system.
Depending on the size of the subdirectories in question and how the
storage is accessed, this might cause noticeable delays and even
freezing.
This commit adds a new option under "View -> Content Display" that
enables users to set "Folder size:" to "No size", completely disabling
directory size counting. Directory size counting is still enabled by
default.
As a third option for "Folder size" is added, the DirectorySizeCount
boolean setting is replaced with a DirectorySizeMode enum setting. The
old setting is migrated using a kconf_update script.
FEATURE: 477187
GUI:
|
|
The bug fix 8f043b2958477d3fe2ef094b7e42f792f4cf0b02 introduced a
secondary code path to trigger a resort of items. However, the previous
way to trigger it only required a small bug fix to work and gives us a
couple of optimisations for free. This commit removes the secondary code
path and fixes the primary one.
CCBUG: 473999
|
|
Signed-off-by: Méven Car <[email protected]>
|
|
|
|
After refreshing the view when
size count is selected as the sortRole,
count is 0 at first. When the actual count is loaded,
the sorting is done according to the count being 0.
This can break the sort order and cause view and model
to be out of sync.
Making sure we always resort all items
when the directory size/item count is changed fixes this
BUG:473999
|
|
Fixes a weird bug with the treeview if one has
```
[PreviewSettings]
Plugins=audiothumbnail
```
in their `dolphinrc`. It can be any PreviewSettings plugin, as long as there it at least one.
Check the bug report for more, but in short, the bug is reproducible like this:
```
create folder tree inside a test folder with
mkdir -p 1/A/B/C/D/E/F/G/H/I
mkdir -p 2/A/B/C/D/E/F/G/H/I
mkdir -p 3/A/B/C/D/E/F/G/H/I
make sure you see the test folder, and expand it
expand folder 1, no need to expand all sub folders
expand folder 2
collapse and expand test folder
3 loses its expansion arrow
if all folders are expanded, only 3 keeps its order
if 1 and all its subfolders are expanded, then test is toggled,
folder 2 also loses its expansion arrow
```
Video of the bug:
The path to the bug is: `KFileItemModelRolesUpdater::slotPreviewFailed` -> `KFileItemModelRolesUpdater::applyResolvedRoles` -> `KFileItemModelRolesUpdater::rolesData` -> `KFileItemModelRolesUpdater::startDirectorySizeCounting`.
This causes Dolphin with preview plugins enabled trying to count the directories again, and the data gets reset even if there is something there. This then resets the `isExpandable` flag and that breaks the treeview, causing all kinds of issues, like being able to expand non-directories and makes things out of sync.
BUG:477607
|
|
Capturing an index is unsafe as it could be out of date when this is
processed. The user could have changed directory inserted a file or
changed sorting order.
|
|
|
|
GIT_SILENT
|
|
|
|
|
|
Hide progressInfo for listDir job on the non-recursize path.
BUG: 472912
|
|
QT_MAJOR_VERSION is undefined, the proper thing is QT_VERSION_MAJOR
This causes the plugin query to not find any plugins
Since we only target KF6 we might as well hardcode that
|
|
|
|
* speeds up incremental builds as changes to a header will not always
need the full mocs_compilation.cpp for all the target's headers rebuild,
while having a moc file sourced into a source file only adds minor
extra costs, due to small own code and the used headers usually
already covered by the source file, being for the same class/struct
* seems to not slow down clean builds, due to empty mocs_compilation.cpp
resulting in those quickly processed, while the minor extra cost of the
sourced moc files does not outweigh that in summary.
Measured times actually improved by some percent points.
(ideally CMake would just skip empty mocs_compilation.cpp & its object
file one day)
* enables compiler to see all methods of a class in same compilation unit
to do some sanity checks
* potentially more inlining in general, due to more in the compilation unit
* allows to keep using more forward declarations in the header, as with the
moc code being sourced into the cpp file there definitions can be ensured
and often are already for the needs of the normal class methods
|
|
|
|
This does not move the settings location in files though.
baby step for https://invent.kde.org/system/dolphin/-/issues/36
|
|
improve stopping, improve data caching
Two user visible changes:
* we can see the dir size changing as it is updated.
* This makes the file counting a lot more reactive, when changing directories for instance.
`KDirectoryContentsCounterWorker::walkDir` is not recursive anymore.
The cache is now shared and only a single thread is used to count size recursively.
|
|
CCBUG: 464722
|
|
|
|
|
|
NO_CHANGELOG
|
|
We do not read the metadata anywhere, consequently we can allow empty metadata.
Consumers now don't have to provide a file and we do no spend time parsing the JSON.
Task: https://phabricator.kde.org/T14538
|
|
Two changes:
* Prioritise size counting for visible path
* stop the worker when switching dirs
|
|
This makes the behavior of kernel cifs/nfs mounts the same as with kio's
smb://. It also helps reduce the chance of the UI hanging due to
opendir/readdir slowing down stat calls on the UI thread.
Signed-off-by: Andrew Gunnerson <[email protected]>
|
|
|
|
|
|
|
|
|
|
If this role isn't in the model yet, it would be `QVariant::Invalid`
which is obviously a distinct type from an empty string list.
This means `KFileItemModel::setData` treats them non-equal and
potentially does expensive operations on the model, which is
called every time a role is resolved in `KFileItemModelRolesUpdater`.
With this change, the number of pointless layout calculations is
significantly reduced.
|
|
This way we get a build time warning if the var isn't defined at all, e.g.
a missing check_include_files() CMake call.
|
|
|
|
Currently, Dolphin doesn't have the ability to sort by file extension (sorting by type means that files will be sorted by mimetype but not by extension). This MR fixes this shortcoming.
BUG: 429579
|
|
With this change (plus https://invent.kde.org/frameworks/kio/-/merge_requests/702), slow files will be treated as remote files for generating previews. This should make browsing local mounts of remote locations smoother.
|
|
|
|
Folder thumbnails are not cached as we cannot know whether something
inside a folder changed. Moreover, they are also a collage of several
files and might even traverse into subdirectories. This takes time.
|
|
This shows a slideshow of thumbs when the user hovers a file item.
|
|
content count
It concerns the detail view.
It can cause multiple preview to be generated for the same file.
This is visible for directories whose icons use a random value for their inner files thumbnails.
Only the view needs to be updated for this model change.
|
|
|
|
QSet<T>::erase(it) is the save way of erasing from a QSet while iterating
it. This saves creating a copy of the QSet, lookup of a node by value in
`QSet<T>::remove(item)`, and repeated rehashing caused by `remove`.
|
|
Reserve sufficient space to avoid reallocation later. Overallocating
actually requires less memory than growing and copying it, and it
is freed immediately after anyway.
|
|
The code below always adds the visible range, up to 5 pages before and
after the visible range (each capped to ResolveAllItemsLimit/2) and
the very first and last page (each up to m_maximumVisibleItems).
Use this number to avoid growing the list later.
|
|
In case the last and/or first page (a page being n=m_maximumVisibleItems
items) overlapped the already inserted items, one item for each range
would be inserted a second time, e.g. for the first page and last page:
`beginExtendedVisibleRange == 0` => `endFirstPage := 0` => `0 <= 0` is
inserted again.
`count == 100; endExtendedVisibleRange == 99; m_maximumVisibleItems == 20`
=> `beginLastPage := 99` => `99 < 100` is inserted again.
|
|
KFileItem::isSlow uses hardcoded logic that returns true if the
filesystem is Smb or Nfs. However,
KFileSystemType::determineFileSystemTypeImpl on Linux returns Nfs also
if the file system in question is a FUSE file system, which includes
NTFS, sshfs, and others.
BUG: 430778
BUG: 431106
|
|
also fixes 1 byte error in size calculation for all folders
|
|
Particularly concerned are .desktop files.
See also:
15baa93640bbb49162f26d439e006729ee9c3441
BUG: 429113
|
|
|
|
This effectively reverts the mutable iterations approach on
2448f88c5f42d7a2040fcf3bcd3c5f2a2f62cd03, and fix crashes
and ghost items when using the filter bar
BUG: 428374
|
|
KFileItemListView contents are periodically scanned by KFileItemModelRolesUpdater.
It uses then KDirectoryContentsCounter to scan directories to determine their size possibly recursively.
Introduce a scanDirectories setting to disable directory scanning by KFileItemModelRolesUpdater.
BUG: 426617
FIXED-IN: 20.08.3
|
|
|
