AMTs fix on random gdx if AMT they are based on failed

[0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q]

https://gitlab.pik-potsdam.de/REMIND/testing_suite/-/blob/a72eb4625634577654cee6f5eaf719c6bdd8a468/README.md#L31-33

SSP1-NPi-AMT_2024-02-02_22.09.55                               nash         Abort 11R*5*Infes   5/100             666656666666           6: Intermed Infes    no   no 
SSP1-PkBudg1050-AMT_2024-02-03_14.49.18         4 hours        nash         Normal completion   34/100            converged              2: Locally Optimal   yes  yes
SSP1-PkBudg650-AMT_2024-02-03_14.49.10                         nash         full.log missing    NA                NA                     NA                   no   no 

SSP1-NPi failed, but SSP1-PkBudg1050 started from the .gdx file from the calibration

/p/projects/remind/modeltests/remind (develop|?? M)$ md5sum ./config/gdx-files/indu_subsectors-buil_simple-tran_edge_esm-POP_pop_SSP1-GDP_gdp_SSP1-En_gdp_SSP1-Kap_debt_limit-Reg_62eff8f7.gdx ./output/SSP1-NPi-AMT_2024-02-02_22.09.55/input.gdx ./output/SSP1-PkBudg1050-AMT_2024-02-03_14.49.18/input.gdx
808fa1380d6b52c44d5cf37ca12830a8  ./config/gdx-files/indu_subsectors-buil_simple-tran_edge_esm-POP_pop_SSP1-GDP_gdp_SSP1-En_gdp_SSP1-Kap_debt_limit-Reg_62eff8f7.gdx
808fa1380d6b52c44d5cf37ca12830a8  ./output/SSP1-NPi-AMT_2024-02-02_22.09.55/input.gdx
808fa1380d6b52c44d5cf37ca12830a8  ./output/SSP1-PkBudg1050-AMT_2024-02-03_14.49.18/input.gdx

which contradicts the documentation and is not really useful (so documentation is OK, behaviour is not)

These columns starting with path… can point to either finished runs or other rows of the current model execution:

• provide a path to an existing gdx file such as ./output/SSP2-Base_2021-12-24_19.30.00/fulldata.gdx. For subfolders of ./output/, writing the folder name SSP2-Base_2021-12-24_19.30.00 is sufficient.
• provide the entry of a title of another row, such as SSP2-Base.
  • If the run with this title is set to start = 1, then runs that point to this title are turned into “subsequent runs“ that will be started after the linked one is finished. Note that in this case, the REMIND code must not change until the full.gms file of the last subsequent run was generated.
  • If the run with this title is set to start = 0 or does not exist in the scenario_config_XYZ.csv file, the function configureCfg from configureCfg.R searches in ./output for folder with that title which contain a fulldata.gdx and whose full.log states *** Status: Normal completion, and then picks the one with the latest date and time in the folder name.
  • If no run is found, the script also looks in cfg$modeltests_folder and so if you use a title of a scenario from the automated model tests in scenario_config_AMT.csv, the most recent of these run is picked which is useful for testing purposes.

[0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q]

Just realised that the AMTs do not have path_gdx defined. So if more a configuration bug.

[0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q]

My original argument is valid.

• SSP1-PkBudg1050 defines path_gdx_bau to be SSP1-NPi,
• SSP1-NPi failed,
• ergo SSP1-PkBudg1050 should not run.

[orichters]

@dklein-pik implemented that runs that do not actually need a path_gdx_bau (because they are not NDC runs) have this dependency removed automatically. So this is not actually an error. #1366 Might be worth cleaning up the config files as well, though.