From 6f9da6dbda271befda85e206a231a8b71e0fb4b7 Mon Sep 17 00:00:00 2001 From: pi-mosaic Date: Tue, 10 Nov 2026 18:41:18 -0400 Subject: [PATCH] hyperloom: build the target `onebit`, not the file `1bit` -- both generators do the wrong thing (engine#1993) The setup instructions told a reader to run cmake --build ~/wt/hl-engine-build -j 4 --target 1bit llama-bench llama-perplexity llama-server `1bit` is the OUTPUT_FILE name (set_target_properties(onebit PROPERTIES OUTPUT_NAME 1bit)), not a target, and the two generators fail differently: Ninja rejects the whole command line ("unknown target '1bit'"), so nothing is built -- not onebit, not the tools; the Makefile generator finds the file `1bit` in the build dir and treats it as an up-to-date file target, so the command exits 0 having built nothing and a stale binary looks fresh. The instructions already pass -G Ninja, so the first is what a reader would have hit. Fixed to `--target onebit` with the reason in two lines, because the same typo reads as a verification failure the next time someone else copies it (engine#1993 reports exactly that: a review rebuilt "main", the import passed, and the binary was 7 hours old). Searched the tree for other call sites: `.github/workflows/bump-hrx.yml:221` already says --target onebit, and no script passes --target 1bit. This was the only site. --- tools/hyperloom/custom/README.md | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/tools/hyperloom/custom/README.md b/tools/hyperloom/custom/README.md index df383a805..b16ef4397 100644 --- a/tools/hyperloom/custom/README.md +++ b/tools/hyperloom/custom/README.md @@ -48,7 +48,14 @@ plus `hl_cli_login.py` and a `launch.sh` wrapper) is gone: upstream now knows th cmake -B ~/wt/hl-engine-build -G Ninja -S ~/wt/hl-engine -DCMAKE_BUILD_TYPE=Release \ -DONEBIT_HRX=ON -DONEBIT_HRX_TOOLCHAIN=/opt/rocm-therock \ -DONEBIT_HRX_LIBHSA=/opt/rocm-therock/lib/python3.14/site-packages/_rocm_sdk_core/lib/libhsa-runtime64.so.1 - cmake --build ~/wt/hl-engine-build -j 4 --target 1bit llama-bench llama-perplexity llama-server + cmake --build ~/wt/hl-engine-build -j 4 --target onebit llama-bench llama-perplexity llama-server + + # `--target onebit`, not `--target 1bit`: `1bit` is the output FILE name + # (set_target_properties(onebit PROPERTIES OUTPUT_NAME 1bit)), not a target. The two + # generators fail differently and both are bad -- Ninja rejects the whole command line + # ("unknown target '1bit'"), so nothing is built at all, while the Makefile generator + # finds a file called `1bit` in the build dir and treats it as an up-to-date file target, + # so the command exits 0 having built nothing and a stale binary looks fresh (engine#1993). ``` The build root is `HL_BUILD` (default `~/wt/hl-engine-build`): `1bit` sits at its top, and the