Tools: unify ryml-parse-emit and ryml-yaml-events, remove ryml-yaml-events - #645
Merged
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #645 +/- ##
==========================================
- Coverage 99.53% 99.52% -0.01%
==========================================
Files 52 51 -1
Lines 14628 14456 -172
==========================================
- Hits 14560 14388 -172
Misses 68 68 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Hancock33
added a commit
to Hancock33/batocera.piboy
that referenced
this pull request
Jul 28, 2026
-------------------------------------------------------------------------------------------------------- batocera-emulationstation.mk 24db6d1510c7ce850d3068dfd54e803ae32cea1e # Version: Commits on Jul 22, 2026 -------------------------------------------------------------------------------------------------------- update po Signed-off-by: Nicolas Adenis-Lamarre <nicolas.adenis.lamarre@gmail.com>, ------------------------------------------------------------------------------------------------ batocera-es-piboy.mk 24db6d1510c7ce850d3068dfd54e803ae32cea1e # Version: Commits on Jul 22, 2026 ------------------------------------------------------------------------------------------------ update po Signed-off-by: Nicolas Adenis-Lamarre <nicolas.adenis.lamarre@gmail.com>, ------------------------------------------------------------------------------------- azahar.mk c711b0ab324d2ec116292b0db7c04a5733b8e559 # Version: Commits on Jul 21, 2026 ------------------------------------------------------------------------------------- BUGFIX: Legacy joystick triggers on linux (#2328) * re-add centering detection to joystick api triggers * patch SDLJoystick to calibrate center at launch, ----------------------------------------------------------------------------------- cemu.mk 50b9e4ba1d4d7cf9821a9cd416378bb94e1ba0ca # Version: Commits on Jul 21, 2026 ----------------------------------------------------------------------------------- vulkan: crash on fence errors instead of possibly spamming the log, ------------------------------------------------------------------------------------------ dolphin-emu.mk b008142f72a099a9bcb971776209504f085ae0be # Version: Commits on Jul 21, 2026 ------------------------------------------------------------------------------------------ Merge pull request #14436 from cbartondock/master Loading custom textures using combination of elf/dol game ids and iso game ids., ---------------------------------------------------- pcsx2.mk v2.7.492 # Version: Commits on Jul 21, 2026 ---------------------------------------------------- - [Qt: Fix Optimal Frame Pacing not inheriting from global](PCSX2/pcsx2#14744) -------------------------------------------------------------------------------------- pcsx2x6.mk 6c2339b647be62b1e26adf9a0ee0a291d0e4df67 # Version: Commits on Jul 22, 2026 -------------------------------------------------------------------------------------- Merge pull request #128 from PS2Homebrew-arcade/fixes-2 Several fixes, ------------------------------------------------------------------------------------ rpcs3.mk 8604f1c5d83fbd256e3e29fdccee4ead805c2689 # Version: Commits on Jul 21, 2026 ------------------------------------------------------------------------------------ cli: allow to boot last savestate of a game, ----------------------------------------------------------------------------------------- xenia-edge.mk b2c8fc8fc338cd0e88dd3007f5bad0db4f8cbbd3 # Version: Commits on Jul 22, 2026 ----------------------------------------------------------------------------------------- [D3D12] Bind shared memory as SRV+UAV on memexport draws This ensures their vertex fetch works fixing Split/Second sky rendering, ----------------------------------------------------------------------------------------- rpi-eeprom.mk bade0f69d5e5add2610fc57db9ca42694443c98b # Version: Commits on Jun 23, 2026 ----------------------------------------------------------------------------------------- rpi-eeprom-digest: Add metadata label for the target-soc Add a metadata field to pieeprom.sig indicating the target-soc. This will allow future versions of the bootloader to skip an incompatible firmware update without reading the contents of the EEPROM image file., ---------------------------------------------------------------------------------------- rpi-utils.mk 6dedfe7cfbe3374aee1aba95ef49b169ee28a9ce # Version: Commits on Jul 09, 2026 ---------------------------------------------------------------------------------------- overlaycheck: Avoid a Perl warning Don't leave an undefined value when a bad compatible string is detected. Signed-off-by: Phil Elwell <phil@raspberrypi.com>, ------------------------------------------------------ appstream.mk v1.1.4 # Version: Commits on Jul 22, 2026 ------------------------------------------------------ Release version 1.1.4, ------------------------------------------------------- rapidyaml.mk v0.16.0 # Version: Commits on Jul 22, 2026 ------------------------------------------------------- ## TL;DR This release focuses on API cleaning and tidying, **in preparation of release 1.0**: - improves tracking of nodes causing deserialization errors - adds serialization to/from `Tree` (not just `NodeRef`/`ConstNodeRef`) - deprecates a good number of functions, mostly around serialization and tree/node building - **brings code coverage to 100%** (rounded from >99.5%) - remove the c4core submodule, vendoring-in just the required c4core source code. The API cleanup introduces a number of deprecations, and some other opt-in but recommended changes: see the migration guide in the next section. Finally there are other fixes and improvements: see the [full changelog](#v016changelogsection), below the migration guide. ## Migration guide This is the list of migration changes, linking to the relevant section below: - [Deprecated `.to_val()` method family for tree building](#toval) - Deprecated operators (but define `RYML_WITH_LEGACY_OPERATORS` to avoid the deprecation): - tree building: [deprecate operators `|=` or `=`](#notequal) - serialization: [deprecate operators `<<` or `>>`](#notshift) - Optional but **recommended**: - [user-implemented `read()` should now return `ReadResult`](#readresult) - [user-implemented `read()`/`write()` should now receive `Tree` and id](#treeid) - [do not use `.load()` (former `>>`) inside `read()`](#noloadinread) - [use `.deserialize_child()` inside `read()`](#deserializechild) Be sure to also read through the quickstart diff for the full overview on how the new features work. > [!TIP] > Hint to see what changed in the quickstart: run `git diff v0.15.2...v0.16.0 samples/quickstart.cpp` on an up-to-date repo clone. <a id=\toval\></a> ### Deprecate `.to_val()` family For tree building, methods `.to_*()` were deprecated in favour of the newly-added `.set_*()` methods. See next section. <a id= otequal\></a> ### Deprecate operators |= and = For tree building, `operator|=` and `operator=` were deprecated in favour of the newly-added `.set_*()` methods: ```c++ c4::yml::Tree tree = ...; c4::yml::id_type node_id = ...; c4::yml::NodeRef node = ...; // before (deprecated) node |= MAP|BLOCK; node[\key\] = \val\; node[\key\] |= VAL_SQUO; node[\seq\] |= SEQ|FLOW; node[\seq2\] |= SEQ; node.append_child().to_seq(\map\); // now, same as above: node.set_map(BLOCK); node[\key\].set_val(\val\, VAL_SQUO); node[\seq\].set_seq(FLOW); node[\seq2\].set_seq(); c4::yml::NodeRef child = node.append_child(); child.set_seq(\map\); // or now, same as above, but from the tree: // more code, but also more efficient tree.set_map(node_id, BLOCK); c4::yml::id_type id_key = tree.append_child(node_id), id_seq = tree.append_child(node_id), id_seq2 = tree.append_child(node_id), id_map = tree.append_child(node_id); tree.set_key(id_key, \key\); tree.set_val(id_key, \val\, VAL_SQUO); tree.set_key(id_seq, \seq\); tree.set_seq(id_seq, FLOW); tree.set_key(id_seq2, \seq2\); tree.set_seq(id_seq2); tree.set_key(id_map, \map\); tree.set_map(id_map); ``` If you prefer to keep on using the legacy operators `=`, `|=`, `<<`, `>>`, you can enable the cmake symbol (or define the macro) `RYML_WITH_LEGACY_OPERATORS`. <a id= otshift\></a> ### Deprecate operators << and >> For serialization, `operator<<` and `operator>>` were deprecated in favour of `.set_serialized()` and `.deserialize()` (which introduce return on error) or `.save()` and `.load()` (which trigger error callback on error): ```c++ c4::yml::Tree tree = ...; c4::yml::id_type node_id = ...; c4::yml::NodeRef node = ...; const T var1 = ...; T var2 = ...; // before (deprecated) node << var1; node >> var2; node << key(var1); node >> key(var2); // now - save() and load() do checks and call error on failure node.save(var1); node.load(&var2); node.save_key(var1); node.load_key(&var2); // or now, without exceptional flow: use .set_serialized() and .deserialize() node.set_serialized(var1); if( ! node.deserialize(&var2)) ... node.set_key_serialized(var1); if( ! node.deserialize_key(&var2)) ... // also now possible from tree! tree.save(node_id, var1); tree.load(node_id, &var2); tree.save_key(node_id, var1); tree.load_key(node_id, &var2); tree.set_serialized(node_id, var1); if( ! tree.deserialize(node_id, &var2)) ... tree.set_key_serialized(node_id, var1); if( ! tree.deserialize_key(node_id, &var2)) ... ``` If you prefer to keep on using the legacy operators `=`, `|=`, `<<`, `>>`, you can enable the cmake variable (or define the macro) `RYML_WITH_LEGACY_OPERATORS`. The rest of the changes are **optional, but recommended**. If you don't do these changes, your code will go on working as before, but the new features and improved behavior will not be available. <a id= eadresult\></a> ### Optional: user-implemented `read()` should now return `ReadResult` Your `read()` functions should now return a `c4::yml::ReadResult`. This enables reporting the exact node on which a deserialization error happens, even if it is nested deep in the tree, which in turn helps tremendously pinpointing problems in YAML source (especially with location tracking): ```c++ // optional: your implementation of read() should be changed from... bool read(c4::yml::ConstNodeRef const& node, T *var) { bool success = ...; return success; } // optional: change to this: c4::yml::ReadResult read(c4::yml::ConstNodeRef const& node, T *var) { bool success = ...; return c4::yml::ReadResult(success, node.id()); } ``` If you don't change, rapidyaml will still report the error, but on the node of the outer-most function returning false, instead of the node at the inner-most function. <a id=\treeid\></a> ### Optional: user-implemented `read()`/`write()` should now receive `Tree` and id If you want to use deserialization from `Tree`, your `read()` and `write()` implementations should now be rewritten for `Tree`, instead of `NodeRef`. This will enable using your `read()`/`write()` functions **with both the `Tree` and `NodeRef` methods**: ```c++ // optional: your implementation of write()/read() should be changed from... void write(c4::yml::NodeRef *node, T const& var) { ... } c4::yml::ReadResult read(c4::yml::ConstNodeRef const& node, T *var) { bool success = ...; return c4::yml::ReadResult(success, node.id()); } // optional: change to this to enable serialization from/to both Tree and NodeRef: void write(c4::yml::Tree* tree, c4::yml::id_type id, T const& var) { ... } c4::yml::ReadResult read(c4::yml::Tree const* tree, c4::yml::id_type id, T *var) { bool success = ...; return c4::yml::ReadResult(success, id); } ``` If you don't change, you will not be able to serialize to/from `Tree`. That is, **having only the node implementation will not work with the `Tree` methods**. You can also provide both functions, and they would be picked up correctly, but that is pointless. In short, **prefer the Tree** version. Note also that **inside the tree version, you can still use the node API**. For example: ```c++ c4::yml::ReadResult read(c4::yml::Tree const* tree, c4::yml::id_type id, T *var) { c4::yml::ConstNodeRef n(tree, id); ... // this way you can keep the function body as before // BUT read next note } ``` <a id= oloadinread\></a> ### Optional: do not use `.load()` (former `>>`) inside `read()` Inside your `write()/read()` implementation **you should now use the new `.deserialize()` calls**. On error, these methods will play nice when they are called from `.deserialize()` on upper-level objects. On the contrary, `.load()` (formerly `>>`) will interrupt execution immediately, and this will prevent those upper-level calls from successfully receiving the actual `ReadResult` and reporting it upwards. If you want to avoid exceptional flow and write code such as ```c++ struct Inner { int foo, bar; }; struct Outer { Inner inner1, inner2; } Outer outer; if( ! node.deserialize(&outer)) ... // we want to enter this branch without triggering an error ``` ... then your `read(tree,id,Inner*)` implementation should not use `.load()`: ```c++ c4::yml::ReadResult read(c4::yml::Tree const* tree, c4::yml::id_type id, Inner *inner) { c4::yml::ReadResult result(tree->is_map(id), id); if(result) result = tree->deserialize_child(id, \foo\, &inner->foo); if(result) result = tree->deserialize_child(id, \bar\, &inner->foo); return result; } // Likewise for the outer type. Note how an Inner error will be // transparently returned: c4::yml::ReadResult read(c4::yml::Tree const* tree, c4::yml::id_type id, Outer *outer) { c4::yml::ReadResult result(tree->is_map(id), id); if(result) result = tree->deserialize_child(id, \inner1\, &outer->inner1); if(result) result = tree->deserialize_child(id, \inner2\, &outer->inner2); return result; } ``` In short, implementing `read()` with `.deserialize()` plays nice with both upper `.deserialize()` and `.load()` calls, whereas implementing `read()` with `.load()` will only work with upper `.load()` calls. <a id=\deserializechild\></a> ### Recommended: use new methods `.deserialize_child()` in `read()` Inside your `read()` implementation, **you should now use the new methods `.deserialize_child()`**. This release also adds a family of methods to simplify `read()` implementations. Each of these returns a `ReadResult` object that is ready to return on error. Consider the following node-based `read()`, already using `.deserialize()`: ```c++ ryml::ReadResult read(ryml::ConstNodeRef const& n, my_type *val) { ryml::ReadResult r(n.is_map(), n.id()); if(r) r = n[\v2\].deserialize(&val->v2); // don't. using [] will throw error if \v2\ is not a child if(r) r = n[\v3\].deserialize(&val->v3); if(r) r = n[\v4\].deserialize(&val->v4); if(r) r = n[\seq\].deserialize(&val->seq); if(r) r = n[\map\].deserialize(&val->map); return r; } ``` Any of the `operator[]` calls (or more precisely the `.deserialize()` calls on its result) may trigger a visit error when no such node exists. To avoid exceptional flow in case of error, you should now use `.deserialize_child()` which simplifies a graceful return with an appropriate error status when the child node does not exist: ```c++ ryml::ReadResult read(ryml::ConstNodeRef const& n, my_type *val) { ryml::ReadResult r(n.is_map(), n.id()); if(r) r = n.deserialize_child(n, \v2\, &val->v2); // there is also a fall-back overload // receiving a default value when // no such child exists if(r) r = n.deserialize_child(n, \v3\, &val->v3); if(r) r = n.deserialize_child(n, \v4\, &val->v4); if(r) r = n.deserialize_child(n, \seq\, &val->seq); if(r) r = n.deserialize_child(n, \map\, &val->map); return r; } ``` Here's the list of new `ReadResult`-returning methods that may be of use in similar scenarios: - `Tree::deserialize_child()`, `ConstNodeRef::deserialize_child()`, `ConstNodeRef::deserialize_child()`: both for keys and indices (on seqs) - `Tree::child_r()`, `ConstNodeRef::child_r()`, `ConstNodeRef::child_r()` - `Tree::find_child_r()`, `ConstNodeRef::find_child_r()`, `ConstNodeRef::find_child_r()` - likewise for sibling: added `sibling_r` and `find_sibling_r()` - The `.get_if()` methods were deprecated in favour of `.deserialize_child()`. <a id=\v016changelogsection\></a> ## Full changelog - Update c4core to [v0.6.0](https://github.com/biojppm/c4core/releases/tag/v0.6.0) - [PR#649](biojppm/rapidyaml#649) commit to ABI stability on patch releases before 0.x, then on minor releases after 1.x - [PR#648](biojppm/rapidyaml#648) improve documentation, add benchmark results - [PR#647](biojppm/rapidyaml#647) remove support for UTF16 and UTF32 encoded files - [PR#646](biojppm/rapidyaml#646) improve Doxygen docs, update yamlscript version - [PR#645](biojppm/rapidyaml#645) tools: unify ryml-parse-emit and ryml-yaml-events, remove ryml-yaml-events - [PR#644](biojppm/rapidyaml#644) amalgamate: add `--fastfloat_sys` to use fastfloat from system - [PR#643](biojppm/rapidyaml#643) reduce source archive size: move large data files used in benchmarks to [rapidyaml-data](https://github.com/biojppm/rapidyaml-data) repo. - [PR#642](biojppm/rapidyaml#642) amalgamate: add option to create single source+header. - [PR#641](biojppm/rapidyaml#641): change uses of `C4_LIKELY()` / `C4_UNLIKELY()` to turn into `[[likely]]` / `[[unlikely]]`. No logic changes. - [PR#640](biojppm/rapidyaml#640): add `.deserialize_child()` methods to simplify `read()` implementations - [PR#639](biojppm/rapidyaml#639): fix names using leading/double underscore (renames only, no logic changes). - [PR#638](biojppm/rapidyaml#638): improve quickstart-ints. - [PR#637](biojppm/rapidyaml#637): add cmake option `RYML_SYSTEM_C4CORE` to consume c4core from `find_package()`. Thanks @uilianries! - [PR#636](biojppm/rapidyaml#636) **remove c4core submodule**, and copy c4core files to rapidyaml (and manage sync): - `ext/c4core.src/`: c4core code used in the rapidyaml library - `ext/c4core.dev/`: c4core code used by tests/benchmarks - `proj/`: project files like cmake files and toolchains - `RYML_STANDALONE` can still be set to `OFF` to compile against an out-of-source c4core (see also the newly added `RYML_SYSTEM_C4CORE`) - the amalgamation tool now adds only the files in `c4core.src` files, and has new option `--c4core-dev` to include the `c4core.dev` files - adds (internal) tools to manage c4core synchronization scenarios. See [ext/README.md](https://github.com/biojppm/rapidyaml/tree/master/ext) for details with this. - [PR#635](biojppm/rapidyaml#635) API cleanup: `NodeType` and `extra::ievt::EventFlags`: - On `Tree` and `NodeRef`, overloads taking `NodeType_e` are now receiving `type_bits`. - This saves operator calls on type queries. - Remove bitwise operators, no longer needed. - Rename `NodeType_e` to `NodeTypeBits` (deprecate `NodeType_e`) - Low impact, no changes should be needed on user code, other than renaming unlikely uses of `NodeType_e`. - Also, for int events: rename `extra::ievt::EventFlags` to `extra::ievt::EventBits` - [PR#620](biojppm/rapidyaml#620) API cleanup: `Tree` and `NodeRef`: - Deprecate `NodeInit` - `Tree` and `NodeRef`: - deprecate `.to_val()` and friends -- add `.set_val()` and friends. - deprecate `operator=(csubstr)` and friends -- use `.set_val()` instead. - deprecate `operator|=(NodeType)` and `operator=(NodeType)` -- use appropriate overload `.set_*(T, NodeType)`. - You can disable compiler deprecation warnings from use of these operators: by enabling the cmake variable (or defining the macro) `RYML_WITH_LEGACY_OPERATORS`. - deprecate `NodeInit` and `NodeScalar` methods in `Tree` and `NodeRef` (use `.set_*()`) - deprecate single-arg `NodeRef::{duplicate,move}(ConstNodeRef)` - deprecate `NodeRef::visit()` and `NodeRef::visit_stacked()` - add `Tree::arena_rem()` - add `RYML_DEFAULT_TREE_ARENA_CAPACITY_START` with default value of 256 - `parse_*()`: internal simplification, no semantic changes - [PR#589](biojppm/rapidyaml#589) API cleanup: serialization and tree building - Refactor serialization code: - Deprecate `operator<<`, use `.load()` / `.load_key()` or `.deserialize()` / `.deserialize_key()` - Deprecate `operator>>`, use `.save()` / `.save_key()` or `.set_serialized()` / `.set_key_serialized()` - Deprecate `key()` tag function and `Key` tag type (needed only for the operators above) - Migration of code triggering serialization: - If you don't want to migrate yet, use `RYML_WITH_LEGACY_OPERATORS` to disable compiler deprecation warnings from use of these operators. - Serialization with `Tree` API is now fully implemented, working exactly the same as `NodeRef` - serialization: - `NodeRef` serialized with `write(NodeRef *, T const&)` and `write_key()` - `Tree` serialized with `write(Tree *, id_type, T const&)` and `write_key()` - No changes in types using `to_chars()` serialization - deserialization: - `ConstNodeRef` deserialized with `read(ConstNodeRef const&, T *)` and `read_key()` - `Tree` deserialized with `read(Tree const*, id_type, T *)` and `read_key()` (removed old `readkey()` approach.) - No changes in types using `from_chars()` serialization - Enables bypass of `NodeRef` serialization to use the tree API, which is faster to compile. - Organize serialization in layers, ensuring use of free functions to enable the user to override at all levels with ADL, for total control of type behavior (explained in the doxygen documentation). - Added many more unit tests to ensure all scenarios are working. - Move scalar specific code to new top-level headers `c4/yml/scalar_charconv.hpp` and `c4/yml/scalar_style.hpp` - Relax deserialization of nodes with `VALNIL` or `KEYNIL`, such that non-fundamental types can now be null-initialized (eg initialize a string from a null scalar). Fundamental types such as ints or floats will still report an error. ```c++ Tree tree = parse_in_arena(\{empty: }\); std::string str; tree[\empty\].load(&str); // relaxed: no longer a deserialization error assert(s.empty()); int val; tree[\empty\].load(&val); // ERROR! as before. ``` - [PR#626](biojppm/rapidyaml#626): ensure accurate reporting of nodes causing deeply nested serialization error. This enables reporting the inner-most node causing the error! To profit from this, `read()` user functions should now return `ReadResult` instead of `bool`. This change is backwards-compatible with legacy `read()` functions returning `bool`, but the legacy version will remain less precise: it will report an inner error on the upmost node returning false, possibly the root if all levels are returning bool. See the new `sample_deserialize_error()` on the quickstart, and the [latest doxygen documentation](https://rapidyaml.readthedocs.io/latest/doxygen/group__doc__serialization__user__types.html). The user functions should change (but not mandatory): ```c++ // before bool read(ryml::ConstNodeRef n, T *var) { ... } // now (should be changed, but not mandatory) ryml::ReadResult read(ryml::ConstNodeRef n, T *var) { ... } ``` - [PR#625](biojppm/rapidyaml#625): Set plain scalar style when serializing arithmetic scalars: improves emit speed of numeric-heavy data payloads. - [PR#616](biojppm/rapidyaml#616) API cleanup: emit - `WriterFile` and `WriterOStream` no longer track the number of emitted bytes. - `error_on_excess` is now used in the emit-to-buffer overloads, and no longer in the main `Emitter::emit_as()` driver function. - [PR#617](biojppm/rapidyaml#617) API cleanup: emit, part 2 - Tidy emit classes among new top-level header files: - `c4/yml/emit_container.hpp`: emit to resizeable contiguous char container (eg `std::string`, `std::vector<char>`) - `c4/yml/emit_buf.hpp`: emit to char buffer (`substr`) - `c4/yml/emit_file.hpp`: emit to C `FILE*` - `c4/yml/emit_ostream.hpp`: emit to STL-like ostreams - The old `c4/yml/emit.hpp` is now a pure umbrella header, including all of the above. For better compilation speed, avoid the umbrella header, and prefer including the concrete header (container, buf, file or stream). - The rest of the emit code was split over these new implementation headers: - `c4/yml/emit_options.hpp`: options to control emitting - `c4/yml/emitter.hpp`: main emitter class - `c4/yml/emitter.def.hpp`: definitions of main emitter class. - `c4/yml/writer_buf.hpp`: policy class to emit to char buffer (`substr`) - `c4/yml/writer_file.hpp`: policy class to emit to C `FILE*` - `c4/yml/writer_ostream.hpp`: policy class to emit to STL-like ostreams - There are no semantic changes: all the `emit_*()` functions remain the same. - Other changes in this PR: - Added `Tree::root_id_maybe()` which is safe to call on an empty tree. - Deprecate `Emitter::max_depth()` - Deprecate `Emitter::options()` setter - [PR#618](biojppm/rapidyaml#618): API cleanup: emit, part 3 - Improve handling of `NaN` and `Inf` in json emitting. - Expose scalar style helpers for json emitting: ```c++ bool scalar_is_plain_number_json(); bool scalar_is_special_json(); bool scalar_is_inf3(); bool scalar_is_nan3(); bool scalar_is_inf_or_nan3(); ``` - Writers: add `C4_ALWAYS_INLINE`. Results in ~10-20% emit improvements. - `file_put_contents()`: add `FILE*` overloads - [PR#621](biojppm/rapidyaml#621) API cleanup: `NodeRef`: - Simplify internal implementation of `{Const}NodeRef::{iterator,children_view}`. - Stop using SFINAE on Node CRTP to distinguish const vs non const, by duplicating the functions in `NodeRef` vs `ConstNodeRef`. No semantic changes. This should improve compilation speed of code containing many node calls. - [PR#622](biojppm/rapidyaml#622) API cleanup: remove preprocess utilities. - [PR#623](biojppm/rapidyaml#623): YAML fuzzing fixes, and close to 1 billion fuzz runs without any errors. These were the only two problems found: - Ensure parse error on multiline keys opening YAML: ```yaml multiline key: value ``` - Fix parse error on Byte Order Mark opening containers: ```yaml <BOM>- - a - ``` - [PR#628](biojppm/rapidyaml#628): Add serialization fuzzing. Relax `c4::atof()` / `c4::atod()`, disable redundant assertions that prevent returning false on bad strings. - [PR#629](biojppm/rapidyaml#629): int events: - Rename `to_chars(substr,DataType)` to `to_str()` (the existing c4 overload would always win because it's an int) - Do not use c4/bitmask.hpp - [PR#633](biojppm/rapidyaml#633): improve coverage to ~99.5%: - Add tests to cover missed lines - Change some errors to assertions; those errors are caught before calling. - `Tree` and `NodeRef`: deprecate `.type_str()`. Use `.type().type_str()` - Tools: add ints parsing to ryml-emit-parse - [PR#634](biojppm/rapidyaml#634): int events: add sample containing ints-only library - [PR#590](biojppm/rapidyaml#590): minor: prefer calling some predicates from `NodeType` ## Thanks - @uilianries , -------------------------------------------------------------------------------------- uwe5622.mk 7bfaaf46ea6834a964672b91b4bf7fd9f0a834e7 # Version: Commits on Jul 22, 2026 -------------------------------------------------------------------------------------- satisfy FORTIFY_SOURCE when enabled by declaring flex arrays, ---------------------------------------------------- box64.mk v0.4.3-4 # Version: Commits on Jul 22, 2026 ---------------------------------------------------- [DYNAREC] Ported the alternate fix from Arm to others (#4091), ------------------------------------------------------------------------------------------ devilutionx.mk ac2d7fc731bc37bc3547eddd2d4c395c868c439f # Version: Commits on Jul 13, 2026 ------------------------------------------------------------------------------------------ Replace libfmt with std::format - `fmt::format` with literal format strings -> `std::format`. - `fmt::format(fmt::runtime(...))` -> `FormatRuntime` in the new `utils/format.hpp`, a std::vformat wrapper that is also safe under the P2905 lvalue requirement of `std::make_format_args`. - `fmt::dynamic_format_arg_store` (Lua logging) -> a small hand-rolled per-field formatter over `std::variant`, returning `std::expected`. - `fmt::format_int` -> `std::to_chars`. - `fmt::join` -> manual joins at the two call sites. - SDL_audiolib: new patch that rewrites aulib_log.h to use std::format and drops the fmt dependency from its CMakeLists. - The `-static-libstdc++` guard now keys off `DEVILUTIONX_STATIC_SDL_AUDIOLIB` instead of `DEVILUTIONX_STATIC_LIBFMT`. - libfmt removed from CMake, vcpkg, Brewfile, CI workflows, Dockerfiles, packaging scripts, UWP project, and docs. - macOS bumped to 13.3+, iOS to 16.3+ (required for `<format>`) - Amiga m68k: provide `__cmpdf2`, which is missing from the hard-float libgcc multilib but referenced via libstdc++'s long double paths. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>, ------------------------------------------------------------------------------------------- jazz2-native.mk f825533bab95a75d5aa7e1ab21cbd067e8c34fb2 # Version: Commits on Jul 21, 2026 ------------------------------------------------------------------------------------------- Fixed build, ---------------------------------------------------------------------------------------------------- sdl2-gamecontrollerdb.mk 15aff6d40aecd704c051aa69132d982eaa39e3b0 # Version: Commits on Jul 22, 2026 ---------------------------------------------------------------------------------------------------- Add mapping for \BigBen Interactive Usb Gamepad\ under Linux (#967) * Add mapping for \BigBen Interactive Usb Gamepad\ under Linux Add mapping for \BigBen Interactive Usb Gamepad\ under Linux created via SDL2 Gamepad tool. Output from lsusb: ``` Bus 002 Device 004: ID 146b:5500 BigBen Interactive Usb Gamepad ``` * Cleanup CRC, name --------- Co-authored-by: Alynne <alviscera@gmail.com>, ------------------------------------------------------- retroarch.mk v1.22.2 # Version: Commits on Nov 20, 2025 ------------------------------------------------------- , ----------------------------------------------------------------------------------------------- retroarch-assets.mk 0959892093bdf85d96206993685a7450b26a1732 # Version: Commits on Jul 21, 2026 ----------------------------------------------------------------------------------------------- Merge pull request #499 from baxysquare/master New & Revised Icons for Automatic & Systematic, ---------------------------------------------------------------------------------------- doomretro.mk db87c60e99b2789a2b21faad929859b552b91914 # Version: Commits on Jul 22, 2026 ---------------------------------------------------------------------------------------- Improve parsing of `toggle` CCMD, -------------------------------------------------------------------------------------- vkquake.mk 36592d4e3629817afe7d31bb1feef4a14ea2955f # Version: Commits on Jul 22, 2026 -------------------------------------------------------------------------------------- Added Quit Prompt option, removed -fitz option (from Ironwail), --------------------------------------------------------------------------------------- vkquake2.mk 0672af0e29a89aead01b09230f27198da435162e # Version: Commits on Jul 22, 2026 --------------------------------------------------------------------------------------- Update README.md, -------------------------------------------------------------------------------------- yquake2.mk 65a8299cad553cedb10b89e8b17c4d34a45ea5e2 # Version: Commits on Jul 21, 2026 -------------------------------------------------------------------------------------- GL3: Significantly reduce draw calls with batching Merge commit '7376ea51' Part of rebase: yquake2/yquake2#1339, ------------------------------------------------------------------------------------------ xash3d-fwgs.mk 2cf8c6363178e58bf22db89a870c49fb763212da # Version: Commits on Jul 22, 2026 ------------------------------------------------------------------------------------------ engine: server: fix a typo, we should set pmove frametime before calling server pmove, instead of server global pmove, like GoldSrc Thanks to @dkenw for drawing attention to this issue., ---------------------------------------------------------------------------------------------- libretro-azahar.mk c711b0ab324d2ec116292b0db7c04a5733b8e559 # Version: Commits on Jul 21, 2026 ---------------------------------------------------------------------------------------------- BUGFIX: Legacy joystick triggers on linux (#2328) * re-add centering detection to joystick api triggers * patch SDLJoystick to calibrate center at launch, -------------------------------------------------------------------------------------------------- libretro-beetle-psx.mk 59e43186cacee4c656a56a4c505b8c16ade90506 # Version: Commits on Jul 22, 2026 -------------------------------------------------------------------------------------------------- psx: honor the cached-code check in the PGXP store handlers The DMA-only invalidation fix added a per-word cached-code bitmap test to lightrec's default store handlers: in that mode, ordinary stores skip invalidation unless they land inside a cached block. The PGXP non-HW handlers in cpu.c were missed - they still skip invalidation unconditionally when DMA-only mode is active: \tif (!psx_dynarec_invalidate) \t\tlightrec_invalidate(state, mem, n); When any PGXP mode is enabled, the RAM/BIOS/scratchpad maps get these ops, and every store routed through the C wrapper resolves to them instead of the checked defaults. The emitted cached-code check does not cover this population either: it is only emitted on the direct-RAM store paths, and guarded, direct and untagged stores all route through the generic wrapper with no emitted check - the measured ~94% of the store mix. Net effect: with PGXP on, DMA-only mode never invalidates on CPU stores at all, and any CPU-side copy of code over cached blocks (e.g. Symphony of the Night streaming a map overlay at a loading door) leaves stale compiled blocks installed. The staleness surfaces as jumps into the previous overlay's code and, eventually, wild accesses into the 0x00800000+ bus hole, which the dynarec treats as fatal. Export lightrec_store_hits_code() and give the four PGXP non-HW store handlers the exact skip condition of their lightrec_default_* counterparts. The unaligned-store handler also inherits the default's 8-byte two-word treatment; it previously invalidated only 4 bytes at the unaligned address, under-invalidating in Full mode as well. Full-mode behavior is otherwise unchanged: the added check only runs when psx_dynarec_invalidate is set., -------------------------------------------------------------------------------------------------- libretro-cannonball.mk 65cb1f5f227db45abcd9d2006efe7687a0d9cb72 # Version: Commits on Jul 21, 2026 -------------------------------------------------------------------------------------------------- engine: lower file-scope integer constants to macros The C++ engine used file-scope \static const\ integer objects as data. In C those are not constant expressions, so any that feed a data table, array size or case label break with \initializer element is not constant\ on the toolchains that don't fold const reads (gcw0/dingux, miyoo, wiiu). The lookup tables were the widest offender: ozoom_lookup.h's ZOOM_LOOKUP[] is built from SIZE1..SIZE5/UNUSED, and many sprite/sound tables likewise. Rather than chase them per-table, lower every file-scope ALL-CAPS scalar integer 'static const' to an object-like macro so it is a genuine integer constant expression everywhere. Exceptions left as objects because they are only ever read at runtime (never in a constant-expression context): CHAN_SIZE (shadowed by a local of the same name in osound.c), VERBOSE, frame_reset, rom_size. omusic.h's multi-declarator HAND_* group becomes a plain enum. No constant among these is address-taken, so none relied on having storage. Verified: full build links with -Wl,--no-undefined, and no file-scope named scalar 'static const' remains in a constant-expression position., -------------------------------------------------------------------------------------------- libretro-gpsp.mk 5b6e751f4abf368509146cd143c949c1946ac1ae # Version: Commits on Jul 21, 2026 -------------------------------------------------------------------------------------------- Delete apply-sound-rate-option.sh, -------------------------------------------------------------------------------------------- libretro-o2em.mk 679d6fec04963f6e70a7ec217e3d0ebb1fe472fc # Version: Commits on Jul 21, 2026 -------------------------------------------------------------------------------------------- libretro-common: restore compat/msvc/stdint.h The libretro-common prune in 82000ac computed its keep-set from the Linux build's textual include closure, which cannot see this file: the pre-2010 MSVC lanes resolve the ordinary '#include <stdint.h>' in every source file through it by appending libretro-common/include/compat/msvc to the INCLUDE path, since those compilers ship no stdint.h of their own. Its deletion broke the windows_msvc2005_x86 build at the first file: src/audio.c(18) : fatal error C1083: Cannot open include file: 'stdint.h': No such file or directory Restored byte-identical from RetroArch @ 45246ce (the same revision the resync came from; upstream's compat/msvc contains only this file). The other five prune deletions were re-audited for the same cross-platform failure mode and are genuinely unreferenced: the string_list.h reference sits behind the unused cdrom path, and audio_resampler.c's stdstring.h uses only header-inline definitions, which the Linux link's --no-undefined already proves. A C90 syntax sweep of the recently touched sources found nothing else the MSVC 2005 frontend would reject beyond its accepted extensions., -------------------------------------------------------------------------------------------- libretro-pc88.mk b5a0e044a914c9a6b8d7b2dd2ddd152f93d35687 # Version: Commits on Jul 22, 2026 -------------------------------------------------------------------------------------------- Merge pull request #79 from cscd98/webos-fix, ---------------------------------------------------------------------------------------------- libretro-prboom.mk c5b09532fa633e015ac75c0a2acfec049f65ded3 # Version: Commits on Jul 21, 2026 ---------------------------------------------------------------------------------------------- README: document the sound system Expand the thin Audio & music highlights and add a dedicated Sound system section covering the reworked pipeline: - output: float32 negotiation with int16 fallback, quantize-once discipline, runtime-selectable 32/44.1/48/96 kHz with in-place SFX retune and song-position resume, deterministic 16.16 frame pacing - music backends: OPL2 at the true 49716 Hz hardware rate through the polyphase Kaiser-windowed-sinc resampler (raw chip sum in, single quantization out), FluidSynth, libmad MP3 at full 28-bit synthesis precision, float-native Ogg, the deliberately integer-only pocketmod MOD decoder, and raw MIDI out; MIDI synth selection vs autodetect - SFX: DMX/WAV/Ogg sources, native-rate storage, 15-bit lerp weight (within 1 LSB of exact), chunked 32-channel mixer, libm-free pitch table - per-backend music position serialize for runahead/rewind, with the render-replay fallback Also corrects the old highlights bullet: dbopl is DOSBox-derived and driven as OPL2 here, not OPL3/Nuked., ---------------------------------------------------------------------------------------------------- libretro-reminiscence.mk e6c0b0039258004f8bc377ddb88c0e931db131ce # Version: Commits on Jul 21, 2026 ---------------------------------------------------------------------------------------------------- ci: revert the ngc dependency override (it broke the whole pipeline) The previous change added 'needs'/'dependencies: build-static-retroarch-ngc' to libretro-build-ngc. That job name is not part of a core's pipeline, so GitLab rejected it with 'undefined dependency: build-static-retroarch-ngc' and refused to run the pipeline at all -- worse than the original failure, which at least let every other platform build. Restore the job to the standard bare form used by every other core (verified identical to libretro-prboom's libretro-build-ngc): just extends .libretro-ngc-static-retroarch-master + .core-defs, with no needs/dependencies. The retroarch-precompiled artifact wiring is the template's responsibility, not the core's; a core must never reference the prepare job by name. Any remaining ngc packaging issue belongs in libretro-infrastructure/ci-templates, not here., ---------------------------------------------------------------------------------------------- libretro-stella.mk 8b36244391a9ba5efe4489b044df39b6edf55738 # Version: Commits on Jul 22, 2026 ---------------------------------------------------------------------------------------------- Fixed a few more AnalogPin::Five/Nine swaps missed in the last commit., ------------------------------------------------------------------------------------------------- libretro-supafaust.mk d6187e5337e6c2646d003db3ab1936727ca75301 # Version: Commits on Jul 22, 2026 ------------------------------------------------------------------------------------------------- Merge pull request #32 from cscd98/android-fix, ---------------------------------------------------------------------------------------------------- libretro-superbroswar.mk 54c6c19e2ed938a072c962359566aa93a7244b88 # Version: Commits on Jul 22, 2026 ---------------------------------------------------------------------------------------------------- Merge pull request #15 from cscd98/android-fix, --------------------------------------------------------------------------------------------- libretro-vba-m.mk e734ed9e6d5813b34d83c6719fe327099f376737 # Version: Commits on Jul 22, 2026 --------------------------------------------------------------------------------------------- translations: transifex pull Signed-off-by: Rafael Kitover <rkitover@gmail.com>, --------------------------------------------------------------------------------------------- libretro-wasm4.mk 4e749b06c3bf75c7511138b2ac4c30a35074f696 # Version: Commits on Jul 22, 2026 --------------------------------------------------------------------------------------------- Merge pull request #870 from cscd98/android-fix vendor - update wasm3,
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.