Roadmap 0.11, Fixes
FX.423 Chip design: OASIS records with flags of 0x80 and up were misread (info-byte)
What was wrong: the OASIS reader and writer (lib/chip/fmt/oasis.js, HD.7) treated each record's info-byte as an unsigned integer. Any record whose flags reach 0x80 (paths with an extension scheme, squares, placements naming their cell, trapezoids) then took one byte too many, so a file from another program stopped partway ("record of unknown kind"). Our own files read back because the writer made the same mistake. Found by KLayout's OASIS of the routed sky130 chip.
What changed: one byte, read and written (Src.byte, Out.byte); the hand-built test file follows the spec; KLayout's files read in full (same shapes as the GDSII). Session chip-11-fu, log docs/0.11-sessions/chip-11-fu.md. Untested: OASIS from a foundry's tools.
Community comments
No comments yet
Add detail as we go: how you would use it, files or steps that show the problem, what other programs do. The team reads every comment.
Sign in or create a free account to add a comment.