FixedBugs.md

 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 4.6.2
+---------------------------
+
+* Perfect foresight simulations of purely backward models could deliver an
+  incorrect result if some exogenous variable appeared with a lag of 2 or more
+  (and neither `block` nor `bytecode` option was used) (fixed via !1759)
+* Perfect foresight simulations of linear models could deliver an incorrect
+  result if the following four conditions were met:
+  + the model was actually declared as linear through the `linear` option
+  + there was an exogenous variable with a lead or a lag
+  + `stack_solve_algo` was equal to 0 (the default) or 7
+  + neither `block` nor `bytecode` option was used
+  (fixed via !1759)
+* In stochastic simulations, for variables that actually do not leave the
+  steady state, reported simulated moments could be spurious (due to division
+  by zero) (#1736)
+* Displayed variance decompositions would take into account measurement error
+  only if measurement errors were present for all observed variables
+  (bug fixed in 6e06acc7)
+* The posterior variance decompositions with measurement errors computed with
+  `moments_varendo` were incorrect (bug fixed in 6e06acc7)
+* `moments_varendo` would not update `oo_.PosteriorTheoreticalMoments` if it
+  was already present from e.g. earlier run of `estimation` (bug fixed in
+  f3329c2d)
+* Identification would in some cases compute wrong Jacobian of moments
+  (bug fixed in !1733)
+* Identification would display incorrect results if parameter dependence was
+  implemented via a steady state file (bug fix in !1732; see discussion in
+  #1733)
+* `generate_trace_plots` would crash when measurement errors were present (bug
+  fixed in f717712e)
+* `estimation` would crash for correlated measurement errors (bug fixed in
+  f717712e)
+* Parallel execution/testing could crash instead of aborting with a proper
+  error message (bug fixed in !1758)
+* Under macOS, Dynare would incorrectly claim that it is compiled for Octave
+  5.2.0 (it is actually compiled for Octave 4.4.1) (bug fixed in
+  b758c154ce6e3a31155308a68e829bedbdbccf66)
+* Using external functions in a model local variable would crash the
+  preprocessor (bug fixed in
+  preprocessor@931935a9420804db7e7d528659ef3be2801886d6)
+* Tolerance criteria for steady state computations were inconsistently set
+  (fixed in 607a273bb0086ed36831b36327c1b3c456e539b2)
+* `stoch_simul` with its default `order=2` would crash with a message about
+  `hessian_eq_zero` not existing if an explicit `order=1` was present somewhere
+  else in the `.mod` file (bug fixed in preprocessor@5cc9f216)
+* Model local variables were not written to the `modfile.json` JSON file (see
+  #1723)
+* Model local variables names would have two spurious underscores when defined
+  in the `dynamic.json` and `static.json` files (but only in the definition,
+  not when they were used, which is inconsistent) (bug fixed in
+  preprocessor@f9ab44da56558ae4be891ad97e52b637484551f6)
+* The `solve_algo=9` option was not accessible. The `solve_algo=10` and
+  `solve_algo=11` options were not accessible with `block` (without `bytecode`)
+  (bug fixed in bd7eb2f88d462fa77fce42faf5b0357018917a22)
+* Under certain circumstances, the `extended_path` command would crash when
+  used in conjunction with the `block` option (see #1717)
+* The `extended_path` command was not working with `bytecode` option (see
+  #1717)
+* The `shock_decomposition` command was not accepting the options of
+  `estimation` related to smoothing (see !1724)
+* The `conditional_forecast` command would display a warning even if the
+  simulation was successful (see !1725)
+* The `prior_trunc` option of `identification` was not working (see !1723)
+* The `proposal_distribution` option of the
+  `tailored_random_block_metropolis_hastings` was not working (see !1727)
+* Perfect foresight simulations of backward models would crash if convergence
+  failed with to complex-valued residuals (see !1726)
+* The diffuse Kalman smoother would crash if `Finf` became singular (see !1728)
+
+
 Bugs fixed in version 4.6.1
 ---------------------------