ip_mreqn.imr_ifindex is not used correctly by the XNU kernel, and
causes us to bind to the default interface; so fallback to ip_mreq
and set the iface source address (not SSM).
Fixes https://gitlab.gnome.org/GNOME/glib/-/issues/3489
...and add ability to select target platform for calling vcvarsall.bat,
so that we can accomodate 32-bit builds and possibly ARM64 builds in the
CI if we need to.
Limit the input size. With a short @find, and a long `init` and `replace`
it’s quite possible to hit OOM. We’re not interested in testing that — it’s
up to the caller of `g_string_replace()` to handle that. 1KB on each of the
inputs should be plenty to find any string parsing or pointer arithmetic
bugs in `g_string_replace()`.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
oss-fuzz#371233785
This patch fixes a build error when compiling with GCC cross compiler for Windows on ARM64.
See issue #3490 for details.
Signed-off-by: Carlo Bramini carlo_bramini@users.sourceforge.netFixes: #3490Closes#3490
This affects the new `g_string_replace()` code which landed on `main` a
few days ago. It does not affect the old implementation of
`g_string_replace()`.
The code for the `f_len == 0` (needle is an empty string) case was
modifying `string` in the loop, without updating any of the string
pointers into it. If the replacement was long enough (or inserted enough
times), this would trigger a realloc of `string->str` and cause all the
string pointers to be dangling.
Fix this by pulling the `f_len == 0` code out into a separate branch and
loop, rather than trying to integrate it into the main loop. This
simplifies the main loop significantly, and makes both easier to verify.
An alternative approach, which doesn’t involve splitting the
`f_len == 0` case out, might have been to track the positions using
indexes rather than string pointers. I think the approach in this commit
is better, though, as it removes the possibility of `f_len == 0`
entirely from the loop, which makes it much easier to verify termination
of the loop.
Add more tests to validate this, including the test from oss-fuzz which
triggered the realloc and found the heap buffer overflow.
The new tests have also been run against the _old_ implementation of
`g_string_replace()` to ensure its behaviour (particularly around `f_len
== 0 && limit > 0`) has not changed.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
oss-fuzz#371043019
This arm of the condition is always true, because 0x00 has been checked
in the previous branch.
This is not going to improve performance, but does mean we now have full
branch coverage of the code via our unit tests, which gives some
assurance that it’s all good.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Helps: #3481
The move to c-utf8 for validation has exposed a few new branches where
our existing (fairly comprehensive) UTF-8 validation test suite didn’t
check things.
Add unit tests for those branches, so we keep code coverage.
I’ve validated (with an independent UTF-8 decoder) that the test vectors
are correctly marked as valid/invalid in the test data (so the tests
aren’t just blindly coded to match the behaviour of the new validator
code).
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Helps: #3481
It turns out it’s not actually been explicitly tested before, even
though it has full code coverage through being called by other code
which is tested.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Scanning for stop chars can require looking through a considerable amount
of input data. In the case there is a single stop character, use memchr()
which can be optimized by the compiler to look at word size or greater
amounts of data at a time.
This moves g_str_is_ascii() from gstrfuncs.c to gutf8.c so that we can
reuse the same SIMD code for ASCII validation.
On Apple Silicon:
Before: 3297 MB/s
After: 26146 MB/s
This is based on the https://github.com/c-util/c-utf8 project and has
been adapted for portability and integration into GLib. c-utf8 is dual
licensed Apache-2.0 and LGPLv2.1+, the latter matching GLib.
Notably, `case 0x01 ... 0x7F:` style switch/case labels have been
converted to if/else which is more portable to non-GCC/Clang platforms
while generating the same assembly, at least on x86_64 with GCC.
Additionally, `__attribute__((aligned(n)))` is used in favor of
`__builtin_assume_aligned(n)` because it is more portable to MSVC's
`__declspec(align(n))` and also generates the same assembly as GCC's
`__builtin_assume_aligned(n)`.
For GCC x86_64 Linux on a Xeon 4214 this improved the throughput of
g_utf8_validate() for ASCII from 750MB/s to around 10,000MB/s (13x).
On GCC aarch64 Linux with an Apple Silicon M2 Pro we go from about
2,200 MB/s to 26,700 MB/s (12x).
Closes: #3481
And improve formatting in a few places while I’m there:
* Add quotes around ‘maybe’ types to make it clearer that ‘maybe’ is
being used as a proper noun
* Add linebreaks so that all doc comments start with a single-sentence
summary of the method
* Improve formatting of constants
* Add a few links to external specifications
See https://developer.gnome.org/documentation/guidelines/devel-docs.html
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Helps: #3250
Instead of using singleton initialized the first time when it's needed
we can create the `DBusProxy` object when needed without blocking since
the interface doesn't have any properties nor signals.
The docs recommend using a small number of `keys` when setting data for an
object, the OpenURI portal uses three different `keys` that can easily be replaced
with `g_task_set_task_data()`.
We can’t quite use `access()` for this, like `g_file_test()` does, as
`g_file_query_info()` considers a broken symlink to exist, so we need to
match that by passing `AT_SYMLINK_NOFOLLOW`.
We also use the `AT_EACCESS` flag, which makes the `faccessat()` call
cheaper on Hurd.
Systems without `faccessat()` will continue to use the
`g_file_query_info()`-based implementation of `g_file_query_exists()`.
(Commit message rewritten by Philip Withnall.)
g_file_query_exists looks like a simpler and faster api than
g_file_query_info. This vfunc lets us actually make it faster,
and avoid allocations in this frequently used code path.
It wasn’t previously clear (to me) whether `end` was returned pointing
to the nul terminator, or to the byte immediately preceding it. Try and
clarify that.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
For the reasons given in
https://gitlab.gnome.org/GNOME/glib/-/merge_requests/4176#note_2233317,
it’s best to not rely on subprocesses when writing tests. Spawning a
subprocess can go wrong, getting feedback and assertion data from a
subprocess is a pain, and making sure the subprocess is killed properly
at the end of the test is hard to get right.
For tests where we are trying to mock a D-Bus service, it’s much more
reliable to run that service in-process (either in the main thread or in
a separate thread).
So, do that for the `fake-document-portal` former subprocess in the
`dbus-appinfo` test: move it to a worker thread.
This speeds the test up, simplifies the build slightly, and should make
the test run more reliable.
In particular, it provides a pattern for future `fake-*-portal` tests to
be built off. This is particularly useful for more complex portals,
where data needs to be relayed back from the mock portal service to the
unit test to check that the code under test has behaved properly. That’s
a pain to do from a subprocess.
Delete the `org.freedesktop.portal.Documents.service` file because we no
longer need to rely on D-Bus service activation within the test, as
we’re setting up the mock service thread explicitly now.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
