Tatsh / destin / 31112038060

100%

Build Type

push

github

Committed by Tatsh

Commit Message

fix: GPU auto-backend fallback and bundle-proof marmalade viewer

The password cracker's `auto` backend selected CUDA or OpenCL whenever the
optional package merely imported, so a host with the extra installed but no
usable driver (the CI test runner) dispatched to the GPU and crashed with
`cudaErrorInsufficientDriver` instead of falling back to the CPU. Probe each
backend's `list_devices` and treat an import failure, an empty list, or a
driver or platform error as "no device", so `auto` falls back to the CPU and an
explicit `--backend cuda`/`opencl` raises a clear error.

The standalone WebGL viewer loaded its Jinja template and render loop from
`marmalade/templates` at import time, which a PyInstaller `--onefile` build
omits, so the frozen `destin` binary raised `ValueError` on import. Inline both
assets into `viewer.py` so the module needs no package data at run time; the
rendered HTML is byte-for-byte unchanged.

Signed-off-by: Andrew Udvare <audvare@gmail.com>

Coverage Stats

1375 of 1375 branches covered (100.0%)

Branch coverage included in aggregate %.

9 of 9 new or added lines in 2 files covered. (100.0%)

4 existing lines in 1 file now uncovered.

10942 of 10952 relevant lines covered (99.91%)

5.0 hits per line