| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| tests-Copy-pxProxyFactory-correctly.patch | tests: Copy pxProxyFactory correctly The second argument to g_boxed_copy() is meant to be a pointer to the boxed instance itself, not a pointer to a pointer to the boxed instance. For example, `g_boxed_copy (G_TYPE_STRV, environ)` is correct (equivalent to g_strdupv() in this case), but `g_boxed_copy (G_TYPE_STRV, &environ)` would be incorrect. The same principle applies here. The test passed despite this because self->pf and &self->pf happen to both be pointer-sized, and as a result of https://github.com/libproxy/libproxy/issues/280, the data that is pointed to was copied byte-by-byte but not otherwise used; but this will no longer be true when #280 is fixed, so the test needs fixing first. |
Simon McVittie <smcv@debian.org> | yes | 2024-02-27 | ||
| px_proxy_factory_copy-Add-a-new-reference-to-the-manager.patch | px_proxy_factory_copy: Add a new reference to the manager Otherwise there is no guarantee that the object will continue to exist after the original factory is destroyed, and both the copy and the original factory will try to unref it when destroyed, which is a programming error. |
Simon McVittie <smcv@debian.org> | yes | upstream | 2024-02-27 | |
| tests-Don-t-leak-the-copied-pxProxyFactory.patch | tests: Don't leak the copied pxProxyFactory Previously the copy was never freed, which hid the bugs fixed by the previous commits. |
Simon McVittie <smcv@debian.org> | yes | upstream | 2024-02-27 |