This is an automated email from the ASF dual-hosted git repository.
xuang7 pushed a commit to branch release/v1.2
in repository https://gitbox.apache.org/repos/asf/texera.git
The following commit(s) were added to refs/heads/release/v1.2 by this push:
new efd41709a8 fix(deps, v1.2): bump torch to 2.12.1 to resolve
CVE-2025-3000 (#6265)
efd41709a8 is described below
commit efd41709a869da0538af7c2ea3543db6599dafd5
Author: Xinyuan Lin <[email protected]>
AuthorDate: Wed Jul 8 10:49:24 2026 -0700
fix(deps, v1.2): bump torch to 2.12.1 to resolve CVE-2025-3000 (#6265)
### What changes were proposed in this PR?
Backport of #6219 to `release/v1.2`.
On `release/v1.2`, **both** torch pins in
`amber/operator-requirements.txt` are at `2.12.0` — the top of the
**CVE-2025-3000** /
[GHSA-rrmf-rvhw-rf47](https://github.com/advisories/GHSA-rrmf-rvhw-rf47)
affected range (`torch.jit.script` memory corruption; CVSS 5.3, `<=
2.12.0`, no upstream patched version). (On `main` only the non-Linux
line was in range, since its Linux line had already moved to
`2.12.1+cpu`.)
| Line | Before | After |
| --- | --- | --- |
| Linux x86_64 (CPU wheel) | `2.12.0+cpu` | `2.12.1+cpu` |
| macOS / Windows / other | `2.12.0` | `2.12.1` |
| `amber/LICENSE-binary-python` | `torch==2.12.0` | `torch==2.12.1` |
Because the advisory's upper bound is `2.12.0`, moving to `2.12.1` takes
both platforms out of range and clears the Dependabot alert on the
release line.
Notes:
- `2.12.1` ships the same cp312 wheel matrix as `2.12.0` (`win_amd64`,
`macosx_*_arm64`, Linux CPU) — a patch-level move with no
platform-coverage change.
- `amber/LICENSE-binary-python` is bumped in lock-step so the
binary-license check stays in sync (no drift).
- `torch.jit.script` is never called in Texera; `torch` is only a
transitive runtime dependency of `transformers` (the HuggingFace
operators), so the vulnerable path is unreachable — this is dependency
hygiene to clear the alert.
### Any related issues, documentation, discussions?
Backport of #6219 (main). Original issue: #6218.
Advisory:
[GHSA-rrmf-rvhw-rf47](https://github.com/advisories/GHSA-rrmf-rvhw-rf47)
/ [CVE-2025-3000](https://nvd.nist.gov/vuln/detail/CVE-2025-3000).
### How was this PR tested?
Dependency-only change; no code paths are altered. Verified:
- `torch==2.12.1` / `2.12.1+cpu` cp312 wheels exist for `win_amd64`,
`macosx_*_arm64`, and the PyTorch Linux CPU index (same matrix as
`2.12.0`), so both pins stay installable on `release/v1.2`.
- `2.12.1 > 2.12.0`, so both lines exit the CVE-2025-3000 `<= 2.12.0`
affected range.
- `amber/LICENSE-binary-python` bumped to `torch==2.12.1` in the same
commit, keeping the `build / pyamber` binary-license check green.
### Was this PR authored or co-authored using generative AI tooling?
Generated-by: Claude Code (Opus 4.8 [1M context])
Co-authored-by: probe <probe@x>
---
amber/LICENSE-binary-python | 2 +-
amber/operator-requirements.txt | 4 ++--
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/amber/LICENSE-binary-python b/amber/LICENSE-binary-python
index 1eb1ef3066..7af04a9f31 100644
--- a/amber/LICENSE-binary-python
+++ b/amber/LICENSE-binary-python
@@ -326,7 +326,7 @@ Python packages:
- sympy==1.14.0
- threadpoolctl==3.6.0
- tifffile==2026.6.1
- - torch==2.12.0
+ - torch==2.12.1
- zstandard==0.25.0
--------------------------------------------------------------------------------
diff --git a/amber/operator-requirements.txt b/amber/operator-requirements.txt
index 99de2e11a3..8eb1e5e24d 100644
--- a/amber/operator-requirements.txt
+++ b/amber/operator-requirements.txt
@@ -23,8 +23,8 @@ pybase64==1.3.2
# Pin torch to the CPU wheel on Linux x86_64 to avoid the NVIDIA CUDA deps.
--extra-index-url https://download.pytorch.org/whl/cpu
-torch==2.12.0+cpu ; platform_system == "Linux" and platform_machine == "x86_64"
-torch==2.12.0 ; platform_system != "Linux" or platform_machine != "x86_64"
+torch==2.12.1+cpu ; platform_system == "Linux" and platform_machine == "x86_64"
+torch==2.12.1 ; platform_system != "Linux" or platform_machine != "x86_64"
scikit-learn==1.5.0
transformers==5.3.0
