Summary
_format_freshness_report in scripts/check_kernel_freshness.py builds the Slack
alert using only result['source_url'], never including result['kernel_dir'].
Problem
Several local kernels share the same upstream. For example, flash-attn2,
flash-attn3, flash-attn4, layer-norm, rotary, and vllm-flash-attn3 all
map to https://github.com/Dao-AILab/flash-attention.
When any are stale, the Slack report prints the same URL multiple times with no
indication of which local directory needs updating:
📊 Kernel Freshness Report - 3 kernel(s) behind upstream
• https://github.com/Dao-AILab/flash-attention: upstream is 12 days newer
• https://github.com/Dao-AILab/flash-attention: upstream is 5 days newer
Affected file
scripts/check_kernel_freshness.py, function _format_freshness_report, line 257.
The kernel_dir key is already populated in each result dict (line 196) — it is
just never used in the output string.
Proposed fix
Before:
items.append(f"• {result['source_url']}: upstream is {result['days_behind']} days newer")
After:
items.append(f"• {result['kernel_dir']} ({result['source_url']}): upstream is {result['days_behind']} days newer")
One-line change. No new dependencies, no behaviour change beyond the string format.
Summary
_format_freshness_reportinscripts/check_kernel_freshness.pybuilds the Slackalert using only
result['source_url'], never includingresult['kernel_dir'].Problem
Several local kernels share the same upstream. For example,
flash-attn2,flash-attn3,flash-attn4,layer-norm,rotary, andvllm-flash-attn3allmap to
https://github.com/Dao-AILab/flash-attention.When any are stale, the Slack report prints the same URL multiple times with no
indication of which local directory needs updating:
Affected file
scripts/check_kernel_freshness.py, function_format_freshness_report, line 257.The
kernel_dirkey is already populated in each result dict (line 196) — it isjust never used in the output string.
Proposed fix
Before:
After:
One-line change. No new dependencies, no behaviour change beyond the string format.