The ref on the timeout source owned by `SendMessageData` was being
dropped just after attaching the source to the main context, leaving it
unowned in that struct. That meant the only ref on the source was held
by the `GMainContext` it was attached to.
This ref was dropped when returning `G_SOURCE_REMOVE` from
`send_message_with_reply_timeout_cb()`. Before that happens,
`send_message_data_deliver_error()` is called, which normally calls
`send_message_with_reply_cleanup()` and destroys the source.
However, if `send_message_data_deliver_error()` is called when the
message has already been delivered, calling
`send_message_with_reply_cleanup()` will be skipped. This leaves the
source pointer in `SendMessageData` dangling, which will cause problems
when `g_source_destroy()` is subsequently called on it.
I’m not sure if it’s possible in practice for this situation to occur,
but the code certainly does nothing to prevent it, and it’s easy enough
to avoid by keeping a strong ref on the source in `SendMessageData`.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
It already implicitly held a strong ref on its `GTask` values, but
didn’t have a free function set so that they would be automatically
unreffed on removal from the map.
This meant that the functions handling removals from the map,
`on_worker_closed()` (via `cancel_method_on_close()`) and
`send_message_with_reply_cleanup()` had to call unref once more than
they would otherwise.
In `send_message_with_reply_cleanup()`, this behaviour depended on
whether it was called with `remove == TRUE`. If not, it was `(transfer
none)` not `(transfer full)`. This led to bugs in its callers.
For example, this led to a direct leak in `cancel_method_on_close()`, as
it needed to remove tasks from `map_method_serial_to_task`, but called
`send_message_with_reply_cleanup(remove = FALSE)` and erroneously didn’t
call unref an additional time.
Try and simplify it all by setting a `GDestroyNotify` on
`map_method_serial_to_task`’s values, and making the refcount handling
of `send_message_with_reply_cleanup()` not be conditional on its
arguments.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
This is equivalent to the current behaviour, but a little clearer in its
meaning.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
The ownership transfers in this code are a bit complex, so adding some
extra documentation and `g_steal_pointer()` calls should hopefully help
clarify things.
This doesn’t introduce any functional changes, just code documentation.
Another drive-by improvement in the quest for #1264.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
The existing comment in the code was correct that `data` is freed when
the task callback is called, because `data` is also pointed to by the
`user_data` for the task, and that’s freed at the end of the callback.
So the existing code was correct to take a copy of `data->task` before
calling `g_task_return_*()`.
After calling `g_task_return_*()`, the existing code unreffed the task
(which is correct), but then didn’t clear the `data->task` pointer,
leaving `data->task` dangling. That could cause a use-after-free or a
double-unref.
Avoid that risk by explicitly clearing `data->task` before calling
`g_task_return_*()`.
After some testing, it turns out this doesn’t actually fix any bugs, but
it’s still a good robustness improvement.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
The generated gir file marks the size parameter as "out" by default. This is wrong in the context of a caller allocated buffer with a given size. Explicitly marking the size parameter as (in) fixes the issue.
As explained in the previous commit, GNU Coreutils 9.2 changes the behaviour
of `du --bytes` to only count regular files and symlinks.
The previous commit makes the test pass with GNU Coreutils >=9.2, but the
machine running the tests may have an older version, or perhaps even a
reimplementation such as uutils. So we can't rely on the size returned by `du`
to be the consistent across systems any more.
However, the plus side of the new behaviour is that the size reported by `du`
/ `G_FILE_MEASURE_APPARENT_SIZE` is now well-defined across filesystems
(as the sum of the sizes of regular files & symlinks), so we can hardcode it.
Fixes: https://gitlab.gnome.org/GNOME/glib/-/issues/2965
Since GNU Coreutils 9.2 (commit 110bcd28386b1f47a4cd876098acb708fdcbbb25),
`du --apparent-size` (including `du --bytes`) no longer counts all kinds of
files (directories, FIFOs, etc.), but only those for which `st_size` in
`struct stat` is defined by POSIX, namely regular files and symlinks
(and also rarely supported memory objects).
This aligns the behaviour of GLib's `G_FILE_MEASURE_APPARENT_SIZE` flag
with the new GNU Coreutils `du` and correct POSIX use.
Note that this may be a breaking change for some uses.
Link: https://lists.gnu.org/archive/html/bug-coreutils/2023-03/msg00007.html
Fixes: https://gitlab.gnome.org/GNOME/glib/-/issues/2965
Warn when the boolean return isn't used, since we may not initialize
the out arguments in the FALSE case. Update all internal callers to
use the return value. They were already safe, but users outside GLib
may not be.
Previously, `-Wl,--export-dynamic` was in `Libs` key of `gmodule-2.0.pc`,
even though `-Wl` is a compiler flag, rather than a linker one.
This caused issues with API reference builds in evolution-data-server,
which passes the output of `pkg-config --libs` through `--ldflags`
argument of `gtkdoc-scan`, which are forwarded unchanged to `ld`:
ld: unrecognized option '-Wl,--export-dynamic'
Let’s move the flag to `Cflags` so that the compiler can deal with it.
https://gitlab.gnome.org/GNOME/evolution-data-server/-/issues/463
Add support for it, but don’t use it anywhere yet — this is an API
addition, but currently doesn’t cause any functional changes. It’ll be
used in the next commit.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The `inotify_init1()` API has supported this flag for a long time
(possibly since it was first introduced, although I haven’t bothered
doing the archaeology).
This saves a syscall when first connecting to inotify.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Both `GPollableInputStream` and `GPollableOutputStream` are dynamic
interfaces, in that their implementation on a class may only be
functional if certain prerequisites are met at runtime. For example,
a `GConverterInputStream` is only pollable if its base stream is
pollable, and that’s determined at runtime rather than compile time.
As such, both interfaces have a `can_poll()` method. If that method
returns `FALSE`, the behaviour of all other methods on the interface is
undefined.
That was mentioned in the documentation for `can_poll()`, but not any of
the other documentation for the interfaces, which made it a bit hard to
find.
Mention it more widely.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2952
If the libc and kernel support `SOCK_NONBLOCK`, we can specify that in
the `socket()` flags, and avoid a subsequent call to `fcntl()` to set
`O_NONBLOCK`.
For modern Linux distributions, this will save a syscall when creating a
socket.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This should split the code up into logical blocks a bit better, and make
it a bit easier to see what the test is doing at a glance.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This makes the exit conditions for each main loop clearer, and
eliminates use of global variables. It introduces no functional changes.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
If the first part of the test takes less than 3s (which is normal), the
timeout for it is not removed, and could spuriously fire during the
second part of the test, causing a false failure.
Instead of relying on source IDs, just use (and explicitly destroy) a
`GSource` for the timeouts.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The test thought that calling `g_app_info_get()` was a bit of a hack,
but actually it (or calling another `g_app_info_*()` function) is the
right way to use the `GAppInfoMonitor` API.
See the documentation improvements a couple of commits back for details.
The remaining FIXME higher up in the test should probably be fixed by
getting `g_app_info_monitor_get()` to arm the signal. That requires
changes in `g_app_info_monitor_get()` to call `desktop_file_dir_init()`.
That will have to happen another time.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #799
This is put together through git archaeology:
```
git log gio/tests/appmonitor.c
```
The following commits were too trivial to have meaningful copyright:
- 54047080e9
- 4e7d22e268
- f2c1cfe8c7
- f8f344923e
- 3ce00b29ec
- 3468369625
- e9d9edde82
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1415
This partially reverts ed8e86a7d4.
The change to add the criticals (commit ed8e86a7d4) is correct, but
landed too late in the cycle. Let’s downgrade the criticals to debugs
for now, to stop applications seeing a lot of new criticals in their
output. Those criticals are particularly disruptive for command line
applications and unit tests.
Early in the next cycle, the debugs will be re-upgraded to criticals.
This will give applications a whole additional cycle to fix their
ambiguous use of API.
It turned out that a lot of applications have latent bugs around
calling `g_file_info_get_*()` without checking whether an attribute
is set first, and were hence relying on the ‘unknown’ return value
also being an appropriate default for them.
This was compounded by the fact that several non-local GVFS backends
were not setting `GFileInfo` attributes all the time, which caused the
‘missing attribute’ code path to be hit more frequently. For example,
they would only call `g_file_info_set_is_hidden()` with a true value and
never bother with a false one.
It was further compounded by the fact that, while this change landed for
the 2.75.4 release, there did not seem to be extensive integration
testing of that release, and distributions and downstreams went straight
to 2.76.0. That meant we missed the window between 2.75.4 and 2.76.0 to
change, fix or revert this behaviour. GLib relies on distros and
downstreams doing integration testing of unstable releases. We test with
downstream GNOME as part of gnome-build-meta, but do not have the
resources to do integration testing for everybody.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
See: #2907
See: #2932
See: #2934
See: #2945
See: #2948
This is written in pseudocode C which omits all the callback boilerplate
for the async calls. This should hopefully make the overall structure of
the loop more obvious.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #352
As suggested on #352 by Owen Taylor (commit put together by Philip
Withnall, but in Owen’s name as it’s his wording).
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #352
Before commit ed8e86a7d4, this function would have silently returned a
zero-valued `GTimeVal` if the correct attributes weren’t present.
That partially regressed in commit ed8e86a7d4, which made it return with
a critical warning, but without zeroing the `GTimeVal`. The critical
warning can be ignored by users (it doesn’t abort the process unless
`G_DEBUG=fatal-criticals` is set), but the change in behaviour of
zeroing the `GTimeVal` could cause bugs.
See: #2907
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
`gio info` currently prints ‘display name:’ twice. One of those should
be ‘edit name:’ — this regressed in commit
a374b7c806.
Various translations still have the old string, but some unfortunately
do not, so this is effectively a new translatable string again.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This fixes bug #624696.
Incorporates a more recent change from 1dc774a653 to drop the
`g_type_init()` call, and reformats the indentation a bit.
Fixes: #322
They just listed built files. Since the move to Meson, these are all
kept in a separate build directory, not the source tree, so don’t need
to be ignored.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
While we can’t check for any events on it, this at least tests that
creating a file monitor works. It should cover the fix from the previous
commit.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: !3241
This should catch regressions in the critical warning fixed in the
previous commit.
The launch has to have several conditions:
- Session bus is running (to avoid the launch happening via the spawn
codepath)
- Use a non-existent D-Bus name (to trigger a launch error)
- Use a launch context (to hit the critical warning code path)
- Not have a startup ID specified in the platform data — this implies
having an empty launch context
- Use an async launch, as that provides an error handling path
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
It’s possible for the startup ID to be `NULL` if one wasn’t provided in
the platform data passed to `launch_uris_with_dbus()`.
Passing `NULL` to `g_app_launch_context_launch_failed()` causes a
critical warning.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The timeout runs for the entire duration of the test, which is a
function that Meson’s test harness already provides for us.
Meson’s timeout can be easily adjusted by a factor to allow for running
tests more slowly under valgrind. The timeout in the code cannot, which
leads to spurious failures like
https://gitlab.gnome.org/GNOME/glib/-/jobs/2645271.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
If the timeout callback was executed, it would remove the timeout
source, leaving the `g_source_remove()` call in the main function with a
dangling source ID.
Fixes commit 73205b8bbd.
Spotted in https://gitlab.gnome.org/GNOME/glib/-/jobs/2645271.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Previously it was only being set on non-Windows platforms. For
consistency, always set it on Windows too.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2934
Following on from #2907, set various boolean attributes if they have
been requested, or are known for sure, and their value is `FALSE`.
Previously the `FALSE` value would have been implicitly returned by the
getter function, but now doing that without the attribute being
explicitly set will trigger a critical warning.
*Don’t* set these attributes if their value is unknown or there was an
error querying it.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2934
The `is-backup` attribute isn't currently set on Windows. It would
be nice to set such basic attributes on all platforms. Let's set
the attribute to `FALSE` there.
Currently, the `G_FILE_ATTRIBUTE_STANDARD_IS_BACKUP` attribute is set
only when its value is `TRUE`. This is wrong with the latest changes as
the `GLib-GIO-CRITICAL **: 00:54:07.260: GFileInfo created without
standard::is-backup` errors are printed now from the
`g_file_info_get_is_backup` function among others. Let's set this
aattribute also when it is `FALSE`.
Related: https://gitlab.gnome.org/GNOME/glib/-/issues/2934
The lost+found dir isn't detected as hidden currently. This is regression
caused by the commit 728ad64b. Let's change the code a bit to be sure that
the lost+found dir is marked as hidden again.
When the `g_file_copy` function is used with files on BTRFS, the
`GLib-GIO-FATAL-CRITICAL: GFileInfo created without standard::size`
error is printed. This is because the `g_file_get_size` function
is used to obtain the file size for the progress callback, but it uses
the wrong `GFileInfo` object that is meant for attributes to be copied
with the file. The file size attribute is missing there obviously. Let's
obtain the file size over the `fstat` call the same way as it is done in
the `splice_stream_with_progress` function to get rid of those errors
and to fix the progress reporting.
This fixes:
| ../glib-2.75.3/gio/tests/cxx.cpp: In function 'int main(int, char**)':
| ../glib-2.75.3/gio/tests/cxx.cpp:61:15: error: missing sentinel in function call [-Werror=format=]
| 61 | g_test_init (&argc, &argv, NULL);
if built with musl libc
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Add a missing steal call in `schedule_method_call()`. This introduces no
functional changes, but documents the ownership transfer more clearly.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2924
This `GDBusMethodInvocation` may be shared across threads, with no
guarantee on the strong ref in one thread outlasting any refs in other
threads — so it needs a ref in this helper struct.
This should fix a use-after-free where the `GDBusMethodInvocation` is
freed from `g_value_unset()` after `g_signal_emit()` returns in
`dispatch_in_thread_func()` in one thread; but then dereferenced again
in `g_source_destroy_internal()` from another thread.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2924
The `GDBusInterfaceSkeleton` is already stored as the source object of
the `GTask` here, with a strong reference.
Storing it again in the task’s data struct is redundant, and makes it
look like the `GDBusInterfaceSkeleton` is being used without holding a
strong reference. (There’s not actually a bug there though: the strong
reference from the `GTask` outlives the data struct, so is sufficient.)
Remove the unnecessary helper struct member to clarify the code a bit.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2924
This introduces no functional changes; just made it while trying to
debug issue #2925.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2925
Otherwise it’s possible for it to hang around in the `GMainContext`
after the “send message” operation has finished. In the best case, this
will cause the `GTask` and `GDBusMessage` to not be freed when the
calling code expects. In the worst case, it could cause use-after-free
problems if it derefs allocations which have since been freed.
I have not seen either of these problems in practice, but it would be
best for the code to eliminate the risk of them altogether by explicitly
destroying the source when the operation is finished.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2925
This doesn’t introduce any functional changes, but should make the code
a little clearer.
Drive-by improvements while trying to debug #1264.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
This introduces no functional changes, but makes it a little clearer how
the ownership of these `GDBusMessage` instances works. The free function
is changed to `g_clear_object()` to avoid the possibility of somehow
using the messages after freeing them.
Basically just some drive-by docs improvements while trying to debug
issue #1264.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1264
This didn’t actually cause any observable bugs, since the structures of
`PropertyData` and `PropertyGetAllData` were equivalent for the members
which the free function touches.
Definitely should be fixed though.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This appears to fix an intermittent failure seen when sending a D-Bus
message with either of a cancellable or a timeout set.
In particular, I can reliably reproduce it with:
```
meson test gdbus-test-codegen-min-required-2-64 --repeat 10000
```
It can be caught easily with asan when reproduced. Tracking down the
location of the refcount mismatch was a little tricky, but was
simplified by replacing a load of `g_object_ref (message)` calls with
`g_dbus_message_copy (message, NULL)` to switch `GDBusMessage` handling
to using copy semantics. This allowed asan to home in on where the
refcount mismatch was happening.
The problem was that `send_message_data_deliver_error()` takes ownership
of the `GTask` passed to it, but the
`send_message_with_replace_cancelled_idle_cb()` and
`send_message_with_reply_timeout_cb()` functions which were calling it,
were not passing in a strong reference as they should have.
Another approach to fixing this would have been to change the transfer
semantics of `send_message_data_deliver_error()` so it was `(transfer
none)` on its `GTask`. That would probably have resulted in cleaner
code, but would have been a lot harder to verify/review the fix, and
easier to inadvertently introduce new bugs.
The fact that the bug was only triggered by the cancellation and timeout
callbacks explains why it was intermittent: these code paths are
typically never hit, but the timeout path may sometimes be hit on a very
slow test run.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #1264
mkstemp-like family of functions also use g_open () under the hood so
they should pass the O_CLOEXEC flag there for race-free setting of the
close-on-exec flag.
setmntent () call uses the same mode flag set as fopen (), so it should
also include the "e" mode flag for race-free setting of the close-on-exec
flag.
All Unix CRTs examined: glibc, musl, BSDs, Apple libc, Android bionic
ignore unknown fopen () mode flags, so this flag can be added
unconditionally for Unix builds.
Only Windows CRT is intolerant of these, so the single case in
g_dbus_address_connect () where the fopen () call is shared between Unix
and Windows needs appropriate platform-specific handling.
Skipped the call sites in libcharset and xdgmime copylibs.
The `equal_func` closure can already have all required information
available without the item, and passing the item via the closure instead
of an explicit parameter is more natural for languages that have a
concept of closures that can capture variables.
Otherwise, the content of the buffer is thrown away when switching
from reading via a GDataInputStream to unbuffered reads when waiting
for the "BEGIN" line.
(The code already tried to protect against over-reading like this by
using unbuffered reads for the last few lines of the auth protocol,
but it might already be too late at that point. The buffer of the
GDataInputStream might already contain the "BEGIN" line for example.)
This matters when connecting a sd-bus client directly to a GDBus
client. A sd-bus client optimistically sends the whole auth
conversation in one go without waiting for intermediate replies. This
is done to improve performance for the many short-lived connections
that are typically made.
This reverts commit 27bee8fe5d.
Inevitably, despite testing the CI multiple times before merging commit
27bee8fe, the CI is now failing again in the `socket` test due to (what
I continue to assume is) the kernel regression:
https://gitlab.gnome.org/martinpitt/glib/-/jobs/2585332
In order to unblock development on `main` expediently, I guess I’ll just
revert the revert.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Reopens: #2879
It’s not meant to be exposed publicly yet (we’re not ready to stabilise
it), but it was incorrectly decorated with `GLIB_AVAILABLE_IN_2_76`.
We can’t remove the decorator and use it that way, as it’s called in
libgio, so we have to expose it using `GLIB_PRIVATE_CALL()`.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2876
Don’t just set them when they’re true and rely on their non-presence
being evaluated to `FALSE`. That means that they erroneously don’t get
returned in `g_file_info_list_attributes()`.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2907
`g_file_info_get_is_hidden()` should not be called without checking the
attribute is set first, just as with the calls higher up in this code.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2907
As documented in a previous commit, these functions should not be called
without the right attributes being present in the `GFileInfo`. Add
critical warnings to make this more obvious.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2907
It doesn’t make sense to (for example) call `g_file_info_get_name()` if
the `GFileInfo` doesn’t contain `G_FILE_ATTRIBUTE_STANDARD_NAME`, given
that building the `GFileInfo` is typically a static process and entirely
under the control of the programmer.
By being this restrictive, we avoid having to return ‘unknown’ values
for some of these standard APIs, particularly the numeric ones such as
`g_file_info_get_size()`. If APIs like that were to work correctly in
the face of a `GFileInfo` without `G_FILE_ATTRIBUTE_STANDARD_SIZE`
specified, they’d have to be able to return a value to indicate the
attribute is missing. Returning `0` or `G_MAXSIZE` to indicate that
would be ambiguous.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2907