The tests currently create various files and directories with generic names
and manipulate with them directly in the `/tmp` directory. The tests also
remove all files in `/tmp` with the generic prefixes like `move`, `copy`,
etc. This can simply cause test suite failures if files with such names
already exist in the `/tmp` directory. Also, the test suite can easily remove
files created by other applications. Let's always create an unique directory
in `/tmp` for tests to fix the mentioned problems...
Mostly the port is straightforward, we connect to tracker-miner-fs
explicitly over D-Bus instead of the centralized tracker-store daemon
we connected to previously.
The search-engine-tracker test is now isolated from the user's real
Tracker index using the `tracker-sandbox` script provided by Tracker,
and it lets tracker-miner-fs index the test file rather than trying
to synthesize the expected database contents.
There are more changes in nautilus-tag-manager.c. Until now, starred
file information was stored in the tracker-miner-fs database. This has
some downsides, firstly the data is deleted if someone runs `tracker
reset --hard`, secondly it isn't possible to do this from inside a
Flatpak sandbox with Tracker 3.0. because the
This commit changes the NautilusTagManager to set up a private
database inside XDG_DATA_HOME/nautilus/tags. This stores the starred
file information. The database is managed with Tracker, which allows us
to continue using the rename-tracking that tracker-miner-fs provides.
The same limitations apply as before that only files in indexed
locations can be starred.
Add new version of FileOperations interface. This versions adds
PlatformData argument to all methods. Currently supported arguments
are parent-handle and timestamp.
This change is necessary for proper focus handling.
https://gitlab.gnome.org/GNOME/nautilus/merge_requests/504
In order for the tests to work in parallel, we need the file/directory
names not to collide.
This patch fixes the namy for the copy-files test, now all file-operations
tests should be able to work in parallel.
This will save some time/thinking when running tests manually.
G_TEST_BUILDDIR and G_TEST_SRCDIR will be useful if
g_test_build_filename() is ever used, since our primary build mode is
outside the source directory, which would break the function.
File utilities operations need to be tested thoroughly everytime
we make changes to make sure they're fine to use further.
This patch introduces an unit test file for the
nautilus_file_selection_equal.
The traditional include guards are not as easy to handle
and require extra thought into the names.
Pragma once is an easier, more contributor friendly approach.
Closes https://gitlab.gnome.org/GNOME/nautilus/issues/294
general: Remove include guards in favor of pragma once
The traditional include guards are not as easy to handle
and require extra thought into the names.
Pragma once is an easier, more contributor friendly approach.
Closes https://gitlab.gnome.org/GNOME/nautilus/issues/294
Some tests require gtk_init, which requires a display and therefore
cannot be part of the tests to be ran in the GitLab CI.
Split those so we can pass an option to meson to only test the
displayless tests for GitLab CI.
This commit does the following:
* Canonicalize the style:
* Use two-space indentations.
* Un-Autotools-ify option names.
* Don’t align arguments, simply increase indentation.
* Don’t add a space before opening parenthesis in calls.
* Remove unused variables.
* Remove unused dependencies.
* Remove config.h.meson.
* Optimize dependencies.
* Use disabler functionality for libselinux dependency, to save lines.
Currently, the process for getting the common prefix of a list of file
names is a tad too greedy:
1. Find the common prefix of all the strings.
2. Strip the extension from the prefix.
3. Strip trailing punctuation.
Step 2 may strip dots if there’s trailing whitespace and step 3 may
strip useful punctuation (e.g. parentheses). This commit reworks the
process as such:
1. Strip the extension from all the file names.
2. Find the common prefix of all the strings.
3. Trim trailing whitespace.
Fixes#174.
When starting the search providers, some provider might finish
before all providers are started, so a wrong value of providers_running
will be used, making Nautilus crash.
To fix this, keep a queue of the started providers and whenever the
value of the finised/running providers is needed, check the status of
each provider.
https://bugzilla.gnome.org/show_bug.cgi?id=785723
In 7840b533e2, `nautilus-directory-provider` extension point was added
to the code paths used by the tests. When glib is asked for extensions,
it assumes the extension point is registered so it segfaults.
This patch registers the extension point before the tests are run.
https://bugzilla.gnome.org/show_bug.cgi?id=779178
Since it’s 2017 already, Nautilus should use a build system that doesn’t
take longer to set up the build than it takes to actually build. An
observed build time using Ninja of roughly one-fifth of what it took
Autotools is more than reason enough to add support for Meson. Along
with that, this commit adds a convenience script to generate a tarball
for releases, since we use libgd as a submodule and Meson does not
handle source distributions.
https://bugzilla.gnome.org/show_bug.cgi?id=778167
This commit removes git.mk and adds hand-written gitignore files. That
is needed to ignore build/, which is the directory of choice for Meson
builds.
https://bugzilla.gnome.org/show_bug.cgi?id=778167
And make the style of Nautilus the same for all files.
Hopefully we can fix all the style issues we can find in the next days,
so expect a little of movement on this.
https://bugzilla.gnome.org/show_bug.cgi?id=770564
The current style of nautilus is rather poor and mixes at least 3
different code styles.
Specific issues that affect the most common contributors to Nautilus
performance are:
- tabs for multiline alignment.
- mix of tabs and spaces.
- errors on no braced one liners conditionals.
- errors on non braced case statements with variable declarations.
So I would say requirements for the style is to address the previous
issues and also be a well known style. I don't want new contributors
to see a new style completely different from C books authors.
So far, I found Allman (aka BSD) style which seems the choice of most C
books authors as far as I can see on internet, and it address the
previous mentioned issues.
Since uncrustify doesn't support the aligment of parameters we do for
multiple stars "**", we also added a script made by Sebastian Wilmet
to align those.
As a matter of practicity I'm going to convert all Nautilus style to
this one, and if the next person who contributes most on Nautilus has
a different choice, please feel free to change it to whatever makes your
performance and your contributors performance the best.
https://bugzilla.gnome.org/show_bug.cgi?id=770564
Build failure in Continuous:
../../test/test-file-utilities-get-common-filename-prefix.c: In function
'test_many_strings':
../../test/test-file-utilities-get-common-filename-prefix.c:366:9:
error: 'for' loop initial declarations are only allowed in C99 or C11
mode
for (int i = 0; i < 500; ++i) {
New Folder with Selection currently doesn't offer a folder name. It
would be better if it suggested a folder name based on the files that
are selected.
With this change, it now looks for a common filename prefix of the
selected files, and pre-populates the folder name entry with that. If no
common prefix is found that is greater than 3 characters long, the
folder name entry will be left blank.
https://bugzilla.gnome.org/show_bug.cgi?id=747907
And fix make distcheck.
Although libnautilus-private seem self contained, it was actually
depending on the files on src/ for dnd.
Not only that, but files in libnautilus-private also were depending on
dnd files, which you can guess it's wrong.
Before the desktop split, this was working because the files were
distributed, but now was a problem since we reestructured the code, and
now nautilus being a library make distcheck stop working.
First solution was try to fix this inter dependency of files, but at
some point I realized that there was no real point on splitting some of
those files, because for example, is perfectly fine for dnd to need to
access the window functions, and it's perfectly fine for the widgets
in the private library to need to access to all dnd functions.
So seems to me the private library of nautilus is somehow an artificial
split, which provides more problems than solutions.
We needed libnautilus-private to have a private library that we could
isolate from extensions, but I don't think it worth given the problems
it provides, and also, this not so good logical split.
Right now, since with the desktop split we created a libnautilus to be
used by the desktop part of nautilus, extensions have access to all
the API of nautilus. We will think in future how this can be handled if
we want.
So for now, merge the libnautilus-private into src, and let's rethink
a better logic to split the code and the private parts of nautilus than
what we had.
Thanks a lot to Rafael Fonseca for helping in get this done.
https://bugzilla.gnome.org/show_bug.cgi?id=765543
It was working as:
- Search directory starts a new search
- Search engine starts all search providers, each one in its
own thread.
- User changes query
- Search engine stops all search providers
- Searchs providers, since they are in its own thread, cancel
in a unknown time.
- Search directory starts a new search, even before all providers
are finished.
- Search engine is marked as need to restart.
- Search providers finished.
- Search engine emits finished signal, since all search providers
now are stopped.
- Clients doesn't have a way to know if the engine
actually finished searching the current search, or the previous
search that the client asked to stop. That might confuse clients
if they ask for results.
- Search engine restart the search providers without noticing the
client, that thinks that the latest search it started was finished
already.
So to fix this confusion, only report that the engine actually finished
if the engine is not going to restart the search providers.
In this way a client can start a batch of consecutive searches without
the risk of getting search finished signals from previous searches.
Clients now will always get the search-finished signal of the
latest search they started.
Now operations reside in the toolbar, in form of a button with
a popover. In this way we avoid to have a nautilus window hanging
around for it.
When no nautilus window is open, the persistence handlers of nautilus
are enough. This use a notification if the server supports it or
a systray icon in case it doesn't.
