This helps to void deadlocks when two processes call interfaces on each
other one of them being org.freedesktop.DBus.ObjectManager.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
The current code is unsafe to use from multiple threads at once.
GIOStream functions like this are supposed to be semi-threadsafe. It's
allowed for them to be called on both a reader thread and a writer
thread at the same time. Of course, it's still tricky and dangerous,
because it's only *really* threadsafe if the handshake has finished,
and API users have no plausible way to know that because the API
does not require performing an explicit handshake operation. But that's
a glib-networking problem. We can at least avoid the most obvious
threadsafety issue here in the API layer.
Note that we'll need to implement the new vfunc in glib-networking for
this to actually work.
Fixes#2393
The documentation for `g_bus_watch_name()` implies that the
`GDestroyNotify` for the user data will be called in the current thread
default `GMainContext`. Currently, it could be called in any thread, as
`client_unref()` can be called in any thread.
Fix that by deferring it to an idle source if `client_unref()` finalises
the `Client` object in a different thread.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
These were missing from the test before the previous commit ported from
`GMainLoop` to `GMainContext`.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
It makes combination exit conditions a lot easier than when using
`g_main_loop_quit()` from different callbacks.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Work around a bug in Meson versions < 0.50.0 where
`--wrap-mode=nodownload` would disable the use of fallback subprojects,
even if they’d already been downloaded, with the message:
```
Cross dependency libpcre found: NO (tried pkgconfig and cmake)
Cross dependency libpcre found: NO (tried pkgconfig)
Not looking for a fallback subproject for the dependency libpcre because:
Use of fallbackdependencies is disabled.
meson.build:1998:2: ERROR: Dependency "libpcre" not found, tried pkgconfig
```
This workaround can be dropped in the near future when we bump our Meson
dependency to something less ancient. We have a self-imposed requirement
to be buildable using what’s packaged in Debian Stable, which is
currently Meson 0.49.2 — but the new Debian Stable release is coming
soon, with an updated Meson.
Using `--force-fallback-for=libpcre` would have been a more constrained
workaround, but unfortunately that argument didn’t exist in Meson 0.49.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #962
No behavior changes. Just reindent the existing code to avoid a GCC
warning spam.
I considered changing the code to not run fflush() on every iteration of
the loop, but I doubt it matters much, so I left it be.
This should maintain equivalent functionality, apart from that now you
have to pass `--force-fallback-for libpcre` to `meson configure` in
order to use the subproject; rather than specifying
`-Dinternal_pcre=true` to use the internal copy.
This also fixes#642, as the wrapdb copy of libpcre is version 8.37.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #962Fixes: #642
Fedora 29 is no longer available, and I need to rebuild these images so
libpcre is pre-downloaded as a subproject.
Update them to Fedora 31, which is what we use in `fedora.Dockerfile`.
This should allow some blobs to be shared in the container repository.
This isn’t the latest Fedora release, but I don’t want to go through the
hassle of updating all the CI images to F33 right now in addition to
updating the subproject caches.
`python-unversioned-command` is needed because the Android NDK calls
`python` without a version number. F29 must have installed this already.
The Android NDK setup appears to be OK with Python 3.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #962
This is basically a contrived test to trigger the `bytes->user_data !=
bytes->data` condition (and none of the earlier short-circuiting
conditions in that statement) in `try_steal_and_unref()`. This gives
100% line and branch coverage for `gbytes.c`.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The first sentence incorrectly said that it checked the type of the
value, and then the second sentence explicitly said it was a programmer
error to give a value of the wrong type.
According to the code, the second sentence is correct.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2425
WebKit wants these private key properties to be readable in order to
implement a deserialization function. Currently they are read-only
because at the time GTlsCertificate was originally designed, the plan
was to support PKCS#11-backed private keys: private keys that are stored
on a smartcard, where the private key is completely unreadable. The
design goal was to support both memory-backed and smartcard-backed
private keys with the same GTlsCertificate API, abstracting away the
implementation differences such that code using GTlsCertificate doesn't
need to know the difference.
The original PKCS#11 implementation was never fully baked and at some
point in the past I deleted it all. It has since been replaced with a
new implementation, including a GTlsCertificate:private-key-pkcs11-uri
property, which is readable. So our current API already exposes the
differences between normal private keys and PKCS#11-backed private keys.
The point of making the private-key and private-key-pem properties
write-only was to avoid exposing this difference.
Do we have to make this API function readable? No, because WebKit could
be just as well served if we were to expose serialize and deserialize
functions instead. But WebKit needs to support serializing and
deserializing the non-private portion of GTlsCertificate with older
versions of GLib anyway, so we can do whatever is nicest for GLib. And I
think making this property readable is nicest, since the original design
reason for it to not be readable is now obsolete. The disadvantage to
this approach is that it's now possible for an application to read the
private-key or private-key-pem property, receive NULL, and think "this
certificate must not have a private key," which would be incorrect if
the private-key-pkcs11-uri property is set. That seems like a minor
risk, but it should be documented.
A reader might think "how would a process terminate without an exit
status?", or equivalently, "what harm would it do if I assume every
termination has an exit status?" without this reminder that termination
with a signal is also reasonably common.
Signed-off-by: Simon McVittie <smcv@collabora.com>
On Unix platforms, wait() and friends yield an integer that encodes
how the process exited. Confusingly, this is usually not the same as
the integer passed to exit() or returned from main(): conceptually it's
an integer encoding of this tagged union:
enum { EXITED, SIGNALLED, ... } tag;
union {
int exit_status; /* if EXITED */
struct {
int terminating_signal;
bool core_dumped;
} terminating_signal; /* if SIGNALLED */
...
} detail;
Meanwhile, on Windows, wait statuses and exit statuses are
interchangeable.
I find that it's clearer what is going on if we are consistent about
referring to the result of wait() as a "wait status", and the value
passed to exit() as an "exit status".
GSubprocess already gets this right: g_subprocess_get_status() returns
the wait status, while g_subprocess_get_exit_status() genuinely returns
the exit status. However, the GSpawn family of APIs has tended to
conflate the two.
Confusingly, g_spawn_check_exit_status() has always checked a wait
status, and it would not be correct to pass an exit status to it; so
let's deprecate it in favour of g_spawn_check_wait_status(), which
does the same thing that g_spawn_check_exit_status() always did.
Code that needs backwards-compatibility with older GLib can use:
#if !GLIB_CHECK_VERSION(2, 69, 0)
#define g_spawn_check_wait_status(x) (g_spawn_check_exit_status (x))
#endif
Signed-off-by: Simon McVittie <smcv@collabora.com>
Confusingly, g_spawn_check_exit_status() takes a wait status, not an
exit status, so passing g_subprocess_get_exit_status() to it is
incorrect (although both encodings happen to use 0 to encode success
and a nonzero value to encode failure, so in practice this probably
had the desired effect).
Signed-off-by: Simon McVittie <smcv@collabora.com>
Following on from the previous commit, some explicit
`g_main_context_wakeup()` calls were missing from the test code which
only uses `GMainContext`.
Add them, and also add some assertions to check that these functions are
being called in the expected thread (as the code comments say).
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The tests in `gdbus-names.c` use a mixture of `GMainLoop` and iterating
a `GMainContext` directly. Some of the helper functions based around the
`OwnNameData` struct use the `loop` `GMainLoop` even when called from
tests like `watch_with_different_context()` which themselves use
`GMainContext` directly.
Thus, it’s possible for the `GMainLoop` to not be running, while the
test is iterating on `g_main_context_iteration()`. In this case,
`g_main_loop_quit()` is a no-op and will not wake up the `GMainContext`.
This causes the test to livelock in around 1 in 1200 test runs.
Fix this by adding an explicit `g_main_context_wakeup()` call after each
`g_main_loop_quit()` call. A more comprehensive fix would be to port all
the tests in this file to iterating `GMainContext` directly, and drop
all the `GMainLoop` usage, but I don’t have time for that right now.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
