leonarski_fandClaude Opus 4.8
2398330e52
Build Packages / build:rpm (ubuntu2404_nocuda) (push) Successful in 13m55s
Build Packages / build:rpm (rocky8_nocuda) (push) Successful in 14m54s
Build Packages / build:rpm (ubuntu2204_nocuda) (push) Successful in 15m1s
Build Packages / build:rpm (rocky9_nocuda) (push) Successful in 15m11s
Build Packages / build:rpm (rocky8) (push) Successful in 15m6s
Build Packages / build:rpm (rocky8_sls9) (push) Successful in 15m16s
Build Packages / build:rpm (rocky9_sls9) (push) Successful in 15m58s
Build Packages / XDS test (JFJoch plugin) (push) Successful in 7m52s
Build Packages / XDS test (durin plugin) (push) Successful in 8m9s
Build Packages / Generate python client (push) Successful in 29s
Build Packages / Create release (push) Skipped
Build Packages / XDS test (neggia plugin) (push) Successful in 8m8s
Build Packages / Build documentation (push) Successful in 59s
Build Packages / build:rpm (ubuntu2404) (push) Successful in 11m17s
Build Packages / build:rpm (ubuntu2204) (push) Successful in 12m12s
Build Packages / DIALS test (push) Successful in 12m20s
Build Packages / build:rpm (rocky9) (push) Successful in 13m38s
Build Packages / Unit tests (push) Successful in 1h0m37s
A serial-crystallography run on a detector with a large converted geometry (JF17T16, modules tiled vertically + horizontally) aborted with "Array out of bounds (Not enough memory to save image)". An indexed still on such a detector predicts/integrates close to the kMaxReflections (10000) cap; at ~170 B per serialized Reflection that is ~1.7 MB of per-image CBOR metadata, which overflowed the fixed 1 MB the buffer slot reserved on top of the image. The serialization guard then threw and cancelled the whole run. - Raise the per-image metadata headroom from 1 MB to 4 MB (GetImageBufferLocationSize). The worst case - 10000 reflections + 2000 spots (API max) + 65534 azimuthal bins - serializes to 2.78 MB, leaving margin while staying negligible next to the multi-MB image slot. - When metadata still does not fit, drop just that frame (log metadata/image/slot sizes + recycle the slot) instead of aborting, in both the FPGA and Lite receivers. - Add a regression test asserting the worst-case metadata fits the headroom. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>