This is a very subtle leak caused interaction between longjmp and compiler optimisation. Specifically, at the point where the setjmp takes places, these pointers have never been reassigned from its nullptr initialization. Without the volatile keyword, the compiler is free to assume that these pointers remain nullptr when the setjmp returns, and therefore to pass nullptr directly to the mj_deleteModel and mj_deleteData calls in the subsequent catch block, without ever reading the actual pointer values.
PiperOrigin-RevId: 466649447
Change-Id: I51e82ec8b566deb144221e11a99d5124a8267de4
The memory handling in the asset loading logic in user_mesh.cc is currently flipped: it frees the asset content buffer when reading from VFS but not from the filesystem. This causes the VFS object to leave these functions in an invalid state, while leaking locally allocated buffers that hold contents read from the filesystem.
PiperOrigin-RevId: 466643390
Change-Id: I162221e019861f23fb61308facd63dbb0422e113
Currently, if these functions fail to allocate space for the buffer or stack, it will exit without freeing any of the memory successfully allocated so far.
PiperOrigin-RevId: 466271030
Change-Id: Id01c115ec976482ac5d28c373b5eb47d77f424be
- Add `azimuth` and `elevation` to `visual.global`.
- Add `mjv_defaultFreeCamera`.
- Use `mjv_defaultFreeCamera` in `simulate`.
PiperOrigin-RevId: 465505244
Change-Id: I61e60f14a990d4e2d220958adb75c325001978ee
Explicitly check that nmocap won't cause overflow when multiplied by 4, to give a nicer error message.
PiperOrigin-RevId: 465380712
Change-Id: I217e3759d55e9761b47db5efd892b8bc811ea8ae
While creating an mjData, any error triggers mj_deleteData. If buffer or stack is not NULL at that point, an uninitialized pointer will be freed.
PiperOrigin-RevId: 465378329
Change-Id: I9be0eef0648e05e3e5f1361da346a5046485cde7
nbody is already checked to be non-negative, but various pieces of code assume it's nonzero (because there's always a world body).
PiperOrigin-RevId: 465339241
Change-Id: I38ce3764074d03bf9a703aa490964c1eb83cc1da
- Adhesion actuators using contact normals as force transmission mechanism.
- Related video: https://youtu.be/HdBue4MUZysCloses#229
PiperOrigin-RevId: 464389367
Change-Id: I9f69b3cd152d957e8f65870d208788463c036a6d
- Mesh inertias can now be computed exactly for well-formed (no holes) non-convex meshes.
- To activate this feature, set `<compiler exactmeshinertia="true">` (defaults to `false`). This default may change in the future.
- Added `<geom shellinertia="true/false">` (defaults to `false`). When true, geom inertia is computed assuming all the mass is concentrated on the surface, and `density` is interpreted as surface density (mass/area). Currently only mesh geoms are supported.
PiperOrigin-RevId: 464368395
Change-Id: I17afd99b9b221c9d24ae951f6e63d5c61ff89820
The breakage was introduced in af12322e9d.
__has_builtin was introduced to GCC in version 10, and the arithmetic overflow builtins were available since version 5.
Fixesdeepmind/mujoco#386.
Tested with GCC 9.
PiperOrigin-RevId: 461463027
Change-Id: I76627539a5c0e7f5617b56a2c9ada9251bceb0f6
- Never modify `d->ctrl`:
- `ctrl` values which are outside their respective `ctrlrange` are clamped internally, but `d->ctrl` itself is unmodified.
- Similarly, if `mjWARN_BADCTRL` is raised (NaNs or huge values), `ctrl` is cleared internally, but `d->ctrl` itself is unmodified.
- `mjWARN_BADCTRL` is no longer raised for `d->ctrl` values that are larger in absolute value than `mjMAXVAL`, if clamping makes them smaller than that.
- Added `mju_clip()` utility function.
- Added missing warning string for `case mjWARN_BADCTRL`.
PiperOrigin-RevId: 459025950
Change-Id: Ifbf974f9c02c3d70afdaa3114018fc81dfd746d3
- Add basic test for keyframes.
- Add missing documentation for keyframe mocap positions and quaternions.
PiperOrigin-RevId: 459021649
Change-Id: I91cf7ecbddc6262e8c72a868eeb82d627f389fb3
mj_makeModel returns NULL if any of the array sizes are negative. Callers of the function need to handle that case.
PiperOrigin-RevId: 457948558
Change-Id: I9a5c8c0d3293ed0a92c5791f00d3ed34c848e024
- Add `qH` and `qHDiagInv` to `mjData` to save factorized modified inertia.
- Add `mj_EulerSkip`, `mj_implicitSkip`, to `engine_forward.c`.
- Using the above functions, implement `mj_stepSkip` in `engine_derivative.c`.
- Add `mjd_stepFD` and `mjd_transitionFD` to `engine_derivative.c` to compute `mj_step` Jacobians.
- Exploit "Skip" functionality for speed.
- Correctly handle quaternion derivatives.
- Handle warmstarts and control limits.
PiperOrigin-RevId: 456584811
Change-Id: Iee8541f11e7b66feb8f431cb102d9bbe65461f79
- Introduced private function `mj_advance()` as single point of state & time advancement.
PiperOrigin-RevId: 456525969
Change-Id: Iae17217e305c7baf07cb5ecd1e1b84b936421097
When using MSAN, mjData->buffer is marked as poisoned, despite being initialized with zeroes.
This is useful for catching computations that accidentally use uninitialized values.
However, when using engine_print, it's OK to assume the buffer is initialized to zeros.
PiperOrigin-RevId: 456085776
Change-Id: I9d85e37fa82958eb54ac413140543a6581aa46ad
This will allow us to add fields that can't be cast to int or double, without breaking print.
PiperOrigin-RevId: 456005913
Change-Id: I1fccfd11466f4745f69172af0ae61be07642927c
The C++ standard library `std::istringstream` is not guaranteed to parse "inf" and "nan" as valid floating point numbers. In our testing, libc++ does this, but libstdc++ and MSVCRT do not.
PiperOrigin-RevId: 455435431
Change-Id: I99c00303b08c2e4e62acfa3ca52a21a3008f4545
Somehow tests load these libs twice, and construct/destruct globals twice.
Can be preproduced even without sanitizers with logging from GlobalModel::~GlobalModel.
themodel does not need to be visible. If it's static each instance will have own a copy.
_mjMap does not need to use string and avoid lifecycle issues at all.
PiperOrigin-RevId: 454678814
Change-Id: I8fca5e1f06cbc2b413beabcece1219445482c0ed
Copybara-Service