When an URI to a symlink is added to the portal, we open it and we send
the FD (of the target) to the portal. This one has no clue about the original
symlink and so it mounts a file that is named like the target.
g_document_portal_add_documents(), however returns a path that contains
the original name and that one is what is sent to the applications when
used via GDesktopAppInfo.
Basically, this is the situation:
- /tmp/symlink -> /tmp/target
- An application is launched to open file:/tmp/symlink
- The portal creates file:/$XDG_RUNTIME_DIR/doc/ID/target
- Gio converts the path to file:/$XDG_RUNTIME_DIR/doc/ID/symlink
Now, since we can't just pass the symlink to the portal without also
changing the logic there, it's just better to do the conversion ourself,
and so, we use the already-opened fd to figure out the real path of the
opened file, and we return a document file URI that uses the target
basename instead
The RFC9110 explanation of HTTP ETags is far superior to
the disjoint ramblings found in (now-obsoleted) RFC2616,
and serves as a much better reference for comparison
with GFile entiity tags. Also, link directly to relevant section.
This commit has the potential to close a thousand Coverity issues.
On generated code from gdbus-codegen, Coverity typically warns on
skeleton->priv->changed_properties_idle_source happening outside
the skeleketon->priv->lock during finalize(), while it's protected
by this mutex in other parts.
Presumably if the object is in finalization, there should be no
other threads poking at it, so the code is actually safe for the
intended use and the warning moot.
To address this, change gdbus-codegen to prefer g_clear_pointer()
with glib >= 2.38 (should be most often the case nowadays) so this
access is reserved to a single line of code, and mark this line of
code with a Coverity code annotation in order to suppress the
warning.
If the build/test machine is slow, heavily-loaded or otherwise
inconvenienced, it might take a few seconds for the signal to be sent
by the subprocess, received by the message bus, re-broadcasted by the
message bus and received by the test code. Wait a few more seconds
before giving up.
If this test is successful, increasing this timeout will not slow it
down: we stop waiting for the signal as soon as we receive it. This will
only make any difference if the test would have failed.
Signed-off-by: Simon McVittie <smcv@debian.org>
We have two things happening in parallel:
1. The GDBus worker thread writes out an AddMatch call to the socket,
the message bus reads that method call from the other end of the
socket, and the message bus responds by adding the match rule
for the signal subscription
2. The main thread forks, and the child execs
gdbus-connection-flush-helper, which sends the signal that we are
expecting; the message bus receives that signal, and broadcasts it
to subscribers, if any
Normally (1.) wins the race because exec'ing a child process is more
time-consuming than IPC, and the test passes.
However, it is possible for (2.) to win the race. If so, we will never
receive the expected signal (because it was received by the message bus
before the AddMatch() method call, so at the time it was received, the
test was not yet a subscriber); the test waits until the timeout and
then gives up, and the test fails.
For whatever reason, Debian's s390x buildd seems to be reliably failing
this test since this week, having previously passed. I don't know what
changed. I can (very rarely) reproduce the race condition described
above on a developer-accessible s390x machine by repeatedly running the
/gdbus/connection/flush test in a loop.
Bug-Debian: https://bugs.debian.org/1115617
Signed-off-by: Simon McVittie <smcv@debian.org>
With these adjustments, building with clang leads to no warnings:
- The "{ NULL }" statement could be replaced with "{ 0 }" to satisfy
clang, but this way it's explicitly filling all fields
- Even though "i" is not read with these g_array_binary_search calls,
it rightfully should be set
IApplicationActivationManager implements the IForegroundTransfer interface,
which lets us specify if the launched app is allowed to create a window
that steals focus (assuming our process has such ability). By default that
setting is disabled, so launched apps appear in the background and do not
get focus.
See XDG activation (Wayland) and the Startup notification specification (X11)
for equivalent functionality on Unix.
Support for focus-stealing can be set in GAppLaunchContext subclasses like
GdkAppLaunchContext, see gdk_app_launch_context_set_timestamp [1].
[1] https://docs.gtk.org/gdk4/method.AppLaunchContext.set_timestamp.html
GdkAppLaunchContext for Wayland / X11 fetch an opaque token
from the compositor.
On Windows, the system doesn't give you any token. Instead,
the token is implicit, and you can pass that to other
processes via dedicated APIs like AllowSetForegroundWindow.
Due to that, we use an internal format for startup-notify-id
that lets us check what to pass to activated apps.
Currently, the `GDesktopAppInfo` constructors return an error if
validity checks for the `Exec=` line, or other lines, fail. However,
they were ignoring the validity checks done at the `GKeyFile` level, for
invalid UTF-8 or string escaping.
It seems consistent to error out for those too, rather than pretending
that the key file line wasn’t set at all (i.e. treating it like `NULL`).
Spotted by Daniel Kondor.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Fixes: #3771
In the case where mimeapps.list is a symlink, gio-issued updates would
overwrite the file, destroying the symlink in the process.
Instead, this approach recursively follows mimeapps.list symlinks
and overwites the contents of the final file instead.
Closes#3579
It’s possible for the dispatch of the timeout source to race with the
finalisation of the `GMemoryMonitorPoll`, given that the timeout is run
in the GLib worker thread.
Protect against that by holding a thread-safe weak ref on the
`GMemoryMonitor` in the callback data.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Fixes: #3758
Try to avoid casting variables to potentially smaller types to fit
defined probes. This can truncate values and lead to wrong results.
Also make sure that signedness matches.
Since GType can be even 128 bit on CHERI architecture, cast all these
various types used based on platform to uintmax_t which SystemTap
properly processes.
The Meson test harness handles that for us.
With a custom timeout, meson test -t is not useful (which is
surprising for users) and interactive debugging sessions may
terminate unexepectedly.
It’s possible for the server communications to finish one main context
iteration before all of the client communications, depending on how the
kernel queues socket connection messages.
Fixes CI failure: https://gitlab.gnome.org/GNOME/glib/-/jobs/5341950
```
GLib-GIO:ERROR:../gio/tests/socket-listener.c:639:test_accept_multi_simultaneously: 'clients[i].result' should not be NULL
not ok /socket-listener/accept/multi-simultaneously - GLib-GIO:ERROR:../gio/tests/socket-listener.c:639:test_accept_multi_simultaneously: 'clients[i].result' should not be NULL
```
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
As the new comments in the code try to explain, this fixes infinite
blocking which could happen when calling
`g_socket_listener_accept_async()` multiple times in parallel, with more
parallel calls than there are pending incoming connections on any of the
`GSocket`s in the `GSocketListener`.
The way `g_socket_listener_accept_async()` works is to create a set of
`GSocketSource`s when it’s called, one for each of the `GSocket`s in the
`GSocketListener`. Those sources are attached to the main context,
polling for `G_IO_IN` (indicating that the socket has a pending incoming
connection to accept).
When one of the socket sources polls ready, `g_socket_accept()` is
called on it, and a new connection is created.
If there are multiple pending `g_socket_listener_accept_async()` calls,
there are correspondingly multiple `GSocketSource` sources for each
`GSocket` in the `GSocketListener`. They will all poll ready in a single
`GMainContext` iteration. The first one to be dispatched will
successfully call `g_socket_accept()`, and subsequent ones to dispatch
will do likewise until there are no more pending incoming connections.
At that point, any remaining socket sources polling ready in that
`GMainContext` iteration will call `g_socket_accept()` on a socket which
is *not* ready to accept, and that will block indefinitely, because
`GSocket` has its own blocking layer on top of `poll()`.
This is not great.
It seems like a better approach would be to disable `GSocket`’s blocking
code, because `GSocketListener` is using `poll()` directly. We only need
one source of poll truth. So, do that.
Unfortunately, that’s complicated by the fact that
`g_socket_listener_add_socket()` allows third party code to provide its
own `GSocket`s to listen on. We probably can’t unilaterally change those
to non-blocking mode, so users of that API will get what they ask for.
That might include blocking indefinitely. I’ve adjusted the
documentation to mention that, at least.
The changes are fairly simple; the accompanying unit test is less
simple. Shrug. It tests for the scenario fixed by this commit, plus the
scenario fixed by the previous commit.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Fixes: #3739
The changes in commit 30ccfac9cf were not quite correct. The code is
structured so that a single reference to a `GTask` (and hence its
`AcceptSocketAsyncData` task data) is shared across the multiple
`GSocketSource`s which are created for a pending `accept_async()` call.
Setting `returned_yet` to true to short-circuit the remaining
`accept_ready()` dispatches in a given `GMainContext` iteration would
have worked, were it not for the fact that the code then immediately
dropped the last reference it had to the `GTask`, potentially freeing
the structure which contained `returned_yet`. Because of the async
nature of `GTask`, the exact timing of finalisation could vary.
This also meant that the other `GSocketSource`s were not destroyed until
an unknown time later.
Improve on that by explicitly destroying the other `GSocketSource`s as
soon as the first one returns an accepted socket. This causes
`GMainContext` to skip dispatching them, even within the same
`GMainContext` iteration. It also means the separate `returned_yet`
member is unnecessary.
This should fix the immediate issue seen in #3739. However, while
testing it I found a further issue which will be fixed in a following
commit, before I add a unit test.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Helps: #3739
