Browser Hang on Multi-Component Update

[gaieges]

Browser Hang on Multi-Component Update

Environment Details

• Gradio Version: 6.7.0
• Browser: [e.g., Chrome/Safari on macOS]
• OS: macOS 15.7.4 (24G517) 36 GB Apple M3 Pro
• Python Version: 3.12.8

Package versions

gradio==6.7.0
gradio-client==2.2.0
numpy==2.4.2
pandas==3.0.1

Symptoms

When a single gr.Button.click event triggers updates to multiple components simultaneously (two Dataframes, and a DownloadButton), the browser tab hangs completely for ~60s or so, on a M3 Pro.

• The Python backend finishes execution very quickly (verified via logs).
• The "processing" animation in Gradio hangs or disappears, and the tab becomes unresponsive for several seconds or indefinitely.
• The second datatable never shows up

Minimal Reproduction

import os
import tempfile

import gradio as gr
import numpy as np
import pandas as pd


def analyze():
    # 1. Summary Markdown
    summary_md = "# 📊 Summary\n- Status: Success"

    # 3. Small Metrics Dataframe
    summary_df = pd.DataFrame(
        {
            "Metric": ["A", "B", "C"],
            "Year 0": ["$100k", "$80k", "$60k"],
            "Year 1": ["$90k", "$72k", "$54k"],
        }
    )

    # 4. CSV File for Download
    fd, csv_path = tempfile.mkstemp(suffix=".csv")
    with os.fdopen(fd, "w") as f:
        pd.DataFrame({"a": [1]}).to_csv(f)

    # 5. Larger Raw Dataframe
    count = 5
    data_df = pd.DataFrame(
        {
            "ID": range(count),
            "Value": np.random.rand(count),
            "URL": ["https://example.com"] * count,
            "Category": ["Equipment"] * count,
        }
    )

    print(summary_df)
    print(data_df)

    return (
        summary_md,
        gr.update(value=summary_df, visible=True),
        gr.update(value=csv_path, visible=True),
        gr.update(value=data_df, visible=True),
    )


with gr.Blocks() as app:
    gr.Markdown("# Gradio Hang Repro")

    with gr.Row():
        with gr.Column(scale=1):
            init_btn = gr.Button("Trigger Hang", variant="primary")
            dl_btn = gr.DownloadButton("Download results", visible=False)

        with gr.Column(scale=3):
            out_md = gr.Markdown("Initial State")
            summary_table = gr.Dataframe(visible=False)

    with gr.Row():
        gr.Markdown("Data Table")
        data_table = gr.Dataframe(visible=False)

    init_btn.click(fn=analyze, outputs=[out_md, summary_table, dl_btn, data_table])

if __name__ == "__main__":
    app.launch(debug=True)

Stack trace in browser

Uncaught (in promise) RangeError: Set maximum size exceeded
    at Set.add (<anonymous>)
    at Batch.ondiscard (index-B0zdmtA4.js:1582:27)
    at svelte_internal_client.js:1235:11
    at update_reaction (index-B0zdmtA4.js:4986:16)
    at update_effect (index-B0zdmtA4.js:5171:18)
    at #traverse_effect_tree (index-B0zdmtA4.js:1361:6)
    at Batch.process (index-B0zdmtA4.js:1281:30)
    at flush_effects (index-B0zdmtA4.js:1698:10)
    at Batch.flush (index-B0zdmtA4.js:1431:4)
    at Batch.revive (index-B0zdmtA4.js:1572:8)

Steps to Reproduce

1. Run python repro_hang.py.
2. Open the URL in a browser.
3. Click "Trigger Hang".
4. Observe the lack of responsiveness or browser "Page Unresponsive" warning.

Have you searched existing issues? 🔎

• I have searched and found no existing issues

Reproduction

above

Screenshot

No response

Logs

(nothing really useful)


  Metric Year 0 Year 1
0      A  $100k   $90k
1      B   $80k   $72k
2      C   $60k   $54k
   ID     Value                  URL   Category
0   0  0.567604  https://example.com  Equipment
1   1  0.913717  https://example.com  Equipment
2   2  0.575840  https://example.com  Equipment
3   3  0.981499  https://example.com  Equipment
4   4  0.831259  https://example.com  Equipment

System Info

above

Severity

Blocking usage of gradio

[freddyaboulton]

I think it may be a dataframe bug

[gaieges]

Yeah that was my guess too but I couldn't get deep enough into the browser internals to learn more ( / didn't have time ).

[gaieges]

Actually, now that I think about it I don't think it's the Dataframe component. If you replace that with a HTML component and just drop in a <table> tag, the problem persists.