Make get_random_string() return a gchar*.
This is part of the effort to migrate string declarations and
allocations from char to gchar, replacing standard C allocators
(malloc/calloc/strdup) with their glib equivalents
(g_malloc/g_new0/g_strdup).
The primary goal is to prevent mismatched allocator bugs. While char and
gchar are binary compatible, mixing their respective deallocators is
dangerous:
* Freeing malloc'd memory with g_free() (or vice versa) can lead to
memory corruption, double-frees, or crashes.
* This is seems to be the case especiall on non Linux platforms or when
using hardened memory allocators where the GLib slice allocator or
wrappers may differ from the system heap.
By standardizing on gchar, we ensure that our automatic cleanup macros
(like auto_gchar) always invoke the correct deallocator (g_free).
Migrate structure allocations from calloc to glibs g_new0 macro to
improve type safety and memory robustness.
* Type Safety: The macro takes the type name directly, ensuring the
allocated size always matches the pointer type.
* Static Analysis: It guarantees a non-NULL return by aborting on
failure, which silences -fanalyzer warnings regarding potential NULL
pointer dereferences.
* Readability: Removes redundant sizeof() calls and is the glib way
Enable gccs static analyzer to detect memory bugs at compile-time. This
validates our __attribute__((__cleanup__)) macros and improves security
when parsing external XMPP data with zero runtime overhead.
Add security hardening via _FORTIFY_SOURCE=2 and switch to -Og to
provide the necessary optimization for these checks while remaining
debugger friendly.
The goal is to catch many buffer overflows and format string errors
at compile-time or runtime.
-Og may occasionally optimize out very short-lived
local variables, making them invisible in the debugger.
So i'm not sure this is a very good idea. But I hope that we will find
bugs earlier and don't even need to debug that often. We might revert
this change later though in case we run into problems too often.
I'm not aware of any other downsides.
Enable stack protection to detect buffer overflows and harden the binary.
This flag adds a canary to stack frames, causing the
program to terminate immediately if stack corruption is detected.
This helps identify memory safety bugs earlier during development by
turning silent corruption into an immediate crash, and provides
protection against stack-smashing exploits at runtime with negligible
performance overhead.
We don't add this generally because we wan't distributions/users to
decide if they want it. But it can help us find problems early on during
development.
Also refactor the function so we pass it the current version
and don't rely on the define inside of it.
This function only allows version format in the style of "x.y.z".
Optimize prof_occurrences by using pointer arithmetic and
g_utf8_next_char instead of calls to g_utf8_offset_to_pointer.
Improve list construction efficiency in prof_occurrences by
using g_slist_prepend and g_slist_reverse instead of appends.
Also fix a slight oversight that could lead to trouble:
When empty search strings (e.g. empty nicks) are passed
it caused matches at every position.
Update the test case to test for empty nicks.
See last commit with comment regarding cons_show().
In this test here we can see that if we pass NULL cons_show() isn't
called. Making it a bit different that other invalid test cases.
It seems like we didn't even check whether the operations succeed or
not.
* `clear_empty()`
Add assertion to confirm a newly created autocomplete object is empty.
* `find_after_create()`:
Add assertion to ensure `autocomplete_complete()` returns NULL when no
entries are present in th object.
The gpgme_set_passphrase_cb function expects a function
pointer of type gpgme_passphrase_cb_t.
The gpgme_passphrase_cb_t typedef specifies a function that returns
gpgme_error_t so return GPG_ERR_NO_ERROR instead of 0.
Fix https://github.com/profanity-im/profanity/issues/1002.
Was already started in 2021 in https://github.com/profanity-im/profanity/pull/1619 but then abandoned it.
I want to take this opportunity to change a couple of other things regarding building:
* The `--plugins` switch is gone. Use `--python-plugins` and `--c-plugins` instead.
* We don't autoenable depending on dependencies being present. Every feature needs to be enabled explicitly. This helps with having deterministic builds.
* Instead of setting `PACKAGE_STATUS="development/release"` in the configure.ac file we will use the default meson `buildtype`.
Autotools will probably stay a while. The next release will contain both with a note for packagers to try to use meson and give us feedback.
Meson/Ninja is significantly faster and the syntax is much easier to read.
The faster build times eventually will also help when waiting for CI results.
There are some things missing still like our `make dist`, `make doublecheck` and such.
Usage example:
```
meson setup build --buildtype=release -Domemo=enabled -Dpgp=enabled
meson compile -C build
```
The changes described above are only on meson. Autotools still behaves the same. The plan is to keep it as is and remove it once we know everything works in meson.
Use disabler() for optional dependencies.
Extract repeated build type checks into is_debug/is_release variables.
Consolidate compiler flags into single add_project_arguments() calls.
Simplify dependency list building (Meson auto-ignores disabled deps).
Streamline platform checks using 'in' operator.
Remove redundant variables (xscrnsaver_found, gtk_version, config_h_inc).
Simplify ncurses and header detection with clearer fallback chains.
Consolidate install_data() calls for files in same directory.
