Troubleshooting#
Symptom |
Check or action |
|---|---|
INI section/key error |
Use a complete template; all general sections are parsed even for static work |
Input file not found |
Resolve paths against the process directory; use absolute paths without surrounding quotes |
Missing solver executable |
Check the active environment and the matching pygrnwang installation |
Numerical DLL error on Windows |
Activate Conda or use |
Multiprocessing bootstrap RuntimeError |
Add a main guard and run from a script |
Fixed-depth CLI does no work |
The configured depth must be greater than zero |
Missing static tensor during dynamics |
Compute matching static results first when |
Correction enabled but not applied |
Set a justified finite |
Full-tensor stress appears scaled by a million |
INI uses MPa; direct Python APIs use Pa |
Static and dynamic amplitudes disagree |
Compare local rigidity × area × slip against supplied moment, then examine STF and baseline settings |
Unexpectedly missing source patches |
Check CSV missing values, slip threshold and STF cutoff |
Source concatenation or grid reshape failure |
Check equal STF lengths, row counts and |
Wrong or repeated dynamic receiver output |
Check collisions in four-decimal per-point names and concurrent writes |
Reused results after changing the model |
Use a fresh library/output directory; cache signatures do not hash all native files |
Excess memory use |
Reduce source–receiver pairs, grid points, time samples or workers; final arrays remain fully assembled |
Optimized CFS not found by a plotter |
Load its |
GIF import failure |
Install the optional |
Small tutorial refuses to run again |
Select a new or empty |
Useful evidence for a bug report#
Record the source revision, versions, platform, exact command, full error
trace, INI and a small input subset. Preserve native solver logs and
green_lib_info.json. Reproduce with one receiver or a small grid when
possible.
Documentation checks establish that the pages build and reference real signatures. They do not establish numerical convergence or validate all backend/receiver combinations.