This page documents the bugs fixed in released versions of Dynare. For bugs fixed in previous versions of Dynare, please read the dedicated frozen page on the [DynareWiki](http://www.dynare.org/DynareWiki/KnownBugs).
Bugs fixed in version 5.3
-------------------------
* The `notmpterms` option of the `dynare` command triggered a crash if the `block` option of the `model` block was used (fixed in preprocessor@3932196ff452eabeae474cecaca78ca65354acf8)
* When the `use_dll` option was passed to the `model` block, the operator `abs` in the `model` block returned an incorrect result (it returned the integer part of the absolute value) (fixed in preprocessor@8cac64c1cbef2b3a55bb539ffbb429ff1f58939a)
* Occbin:
* the piecewise linear Kalman filter (PKF) crash if the model solution could not be computed for a parameter draw (bug fixed in 3c0f8ec6 and eb416186)
* the piecewise linear Kalman filter (PKF) could crash mode finding if an error was encountered (bug fixed in 5defbdeb927a536f983f31d3e102832151a524ec)
* the piecewise linear Kalman filter (PKF) would crash in the one-constraint case if the fixed point algorithm did not converge (bug fixed in 95a15815)
* the smoother could crash due to the initial states being empty (bug fixed in eb6f17e7) and when encountering errors (bug fixed in ce21248e)
* the smoother fields of `oo_` contained wrong results if the piecewise linear Kalman smoother did not converge (bug fixed in 4d603fb3)
* in pathological cases, seemingly periodic solutions could be incorrectly accepted (bug fixed in d0150997)
* Estimation:
*`mh_recover` and `load_mh_file` did not find the saved proposal density and had to rely on the mode-file (bug fixed in df09084a)
*`load_results_after_load_mh` did not find the location of the results-file (bug fixed in 905e7c43)
*`bayesian_irf` together with `loglinear` produced wrong results (bug fixed in ee8dfb76)
* the diffuse Kalman smoother initialization was wrong when the state transition matrix contained a column of zeros (bug fixed in 81ba9524)
* the diffuse Kalman smoother initialization was wrong when the shock covariance matrix was not diagonal (bug fixed in 178e892e)
* Perfect foresight:
* when solving purely forward or backward models with the PATH solver (`solve_algo=10`), `mcp`-tags were ignored (bug fixed in 0225cb23)
*`linear_approximation` ignored the `nocheck`-option (bug fixed in 9301024d)
* in the presence of a steady state file or a `steady_state_model`-block, the contents of the last `initval` or `endval`-block were ignored and replaced by a steady state (bug fixed in cbbbdf4b)
*`identification` and `dynare_sensitivity` did not pass a `graph_format`-option to subsequent other commands (bug fixed in https://git.dynare.org/Dynare/preprocessor/-/commit/3f674c1af925407b366255737d004afcd31116a3)
*`dynare_sensitivity`:
* stability mapping incorrectly imposed a parameter limit of 52 (bug fixed in bb9e45d0)
* prior sampling did not work with user-specified `prior_trunc=0` (bug fixed in 1a46c1dc)
*`dynare++`: the `dynare_simul.m` was broken (bug fixed in 54d95772)
*`model_diagnostics` did not work with `block_trust_region` (`solve_algo=13,14`) (bug fixed in d76455fb)
