Script 'mail_helper' called by obssrc Hello community, here is the log from the commit of package python-comfy-aimdo for openSUSE:Factory checked in at 2026-09-04 12:38:41 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/python-comfy-aimdo (Old) and /work/SRC/openSUSE:Factory/.python-comfy-aimdo.new.1265 (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-comfy-aimdo" Fri Sep 4 12:38:41 2026 rev:4 rq:1375418 version:0.5.0 Changes: -------- --- /work/SRC/openSUSE:Factory/python-comfy-aimdo/python-comfy-aimdo.changes 2026-08-26 19:53:58.532275935 +0200 +++ /work/SRC/openSUSE:Factory/.python-comfy-aimdo.new.1265/python-comfy-aimdo.changes 2026-09-04 12:39:17.912242142 +0200 @@ -1,0 +2,16 @@ +Wed Sep 2 14:43:05 UTC 2026 - Martin Pluskal <[email protected]> + +- Update to 0.5.0: + * New "memory compiler": a native CUDA malloc-graph recorder that + replays an allocation trace instead of re-deciding it per step, + plus VBAR eviction during CUDA graph capture + * Adds the comfy_aimdo.malloc_graph module and control.record(); + both need the native allocator, which this build does not compile + * Purely additive on the Python side -- no name was removed, and + every other module is byte-identical to 0.4.15, so consumers + pinning a >= floor are unaffected + * Ships a tests/ tree, but every case needs a CUDA device, so + %check stays the import/no-accelerator check +- Extend %check to import the new comfy_aimdo.malloc_graph module + +------------------------------------------------------------------- Old: ---- comfy-aimdo-0.4.15.tar.gz New: ---- comfy-aimdo-0.5.0.tar.gz ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ python-comfy-aimdo.spec ++++++ --- /var/tmp/diff_new_pack.Qz3hn7/_old 2026-09-04 12:39:18.806273548 +0200 +++ /var/tmp/diff_new_pack.Qz3hn7/_new 2026-09-04 12:39:18.808273618 +0200 @@ -18,7 +18,7 @@ %{?sle15_python_module_pythons} Name: python-comfy-aimdo -Version: 0.4.15 +Version: 0.5.0 Release: 0 Summary: AI Model Dynamic Offloader for ComfyUI (pure-Python fallback) License: GPL-3.0-only @@ -60,9 +60,11 @@ %python_expand %fdupes %{buildroot}%{$python_sitelib} %check -# No upstream test suite. The native aimdo.so is not built; confirm the -# Python modules import and the loader leaves lib unset. -%python_expand PYTHONPATH=%{buildroot}%{$python_sitelib} $python -B -c "import comfy_aimdo.control, comfy_aimdo.host_buffer, comfy_aimdo.model_mmap, comfy_aimdo.model_vbar, comfy_aimdo.vram_buffer; assert comfy_aimdo.control.lib is None" +# Upstream's tests/ are CUDA-only: each one calls init("cuda") on a real +# device through the native aimdo.so, which this build does not compile. +# Confirm instead that the shipped modules import and the loader leaves +# lib unset (torch.py excluded: it imports torch at module scope). +%python_expand PYTHONPATH=%{buildroot}%{$python_sitelib} $python -B -c "import comfy_aimdo.control, comfy_aimdo.host_buffer, comfy_aimdo.malloc_graph, comfy_aimdo.model_mmap, comfy_aimdo.model_vbar, comfy_aimdo.vram_buffer; assert comfy_aimdo.control.lib is None" # Patch0: without a CUDA or ROCm PyTorch, init() must report failure rather # than guess a vendor, and it must not emit a warning while doing so. cat > test_no_vendor.py <<'EOF' ++++++ comfy-aimdo-0.4.15.tar.gz -> comfy-aimdo-0.5.0.tar.gz ++++++ ++++ 3691 lines of diff (skipped)
