You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 09b2475
Browse filesBrowse the repository at this point in the historyBrowse files
Cut wordlist memory: stream -w, skip visited when flat (#248)
Loading a 5M-line wordlist peaked around 740 MB. Three copies go:
- wordlists are normalized while they are read (wordlist.Preparer,
EachLine, EachLineOf), so the raw lines are never all held;
- a non-recursive scan keeps no dispatcher visited set, since every
caller passes unique entries; only recursion can repeat a name;
- passive sources and the zone transfer add to one candidates set per
target instead of each rebuilding a map of the whole wordlist.
Measured on 5M lines with -simulate -max-queries 1: peak working set
706/742/767 MB before, 576/559/559 MB after, same run time.
Refs #248
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
0 commit comments