mirror of
https://github.com/odin-lang/Odin.git
synced 2026-09-02 02:03:35 +00:00
cc39d05d35
The instruction table was the only conflict, and none of it was a real
disagreement: upstream c9b28d2ec reformatted every row, our six arm64 commits
edited rows under the old formatting, and the two sets are disjoint -- upstream
touched 40 mnemonics, we touched 160, intersection empty. The previous commit
put our rows in their format so the merge had one style to reason about; what
remained merged with all 124 of their new/changed rows and all 399 of ours
present, and nothing dropped.
The arm64 blobs are regenerated here because the merge makes it mandatory, not
optional: upstream added 35 mnemonics, so the enum reached 856 while the
committed encode_runs blob held 821, and ENCODE_RUNS[u16(m)] indexed past its
end -- the test binary aborted with "Index 821 is out of range 0..<821" before
running. That staleness was already on both branches; upstream's tip fails 28
tests and crashes for the same reason, and is green once regenerated.
KNOWN FAILING, PRE-EXISTING AND OURS: 326 passed, 4 failed --
WHILELT 64 table 0xFF20FC10 tests want 0xFFE0FC10
SVE_ST1B_SCATTER_S table 0xE4008000/0xFFA0E000 want 0xE4408000/0xFFE0A000
SVE_ST1W_SCATTER_S table 0xE5008000/0xFFA0E000 want 0xE5408000/0xFFE0A000
SVE_ST1D_SCATTER_D table 0xE5808000/0xFFA0E000 want 0xE5808000/0xFFE0A000
These are not the merge's doing and not upstream's. Our arm64 commits updated
the test expectations to the corrected SVE encodings but the matching
instruction_table.odin rows were never committed -- the row reads 0xFF20FC10 at
every commit including the merge base. The branch looked green only because the
tests read the stale blob, which does carry the fix. Our own pre-merge tree,
regenerated from its own committed table, fails these same four.
So the source is behind the artifact, which is the same class of defect as the
x86 enc_hint staleness, in the opposite direction. The intended values are
recoverable: the old blob and the test expectations agree on them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Riok9vMpkLmo78wsVKJHhz