bgpd: remove unneeded sort of communities in rmap delete

[mjstapp]

When routemaps delete from a community list, don't need to re-sort the community list: the delete code preserves the sorted order.
this was F-079 from the 2026_05_27 batch of reports from Qifan Zhang

[greptile-apps]

Greptile Summary

This PR removes a redundant sort-and-dedup step (*_uniq_sort()) that followed *_list_match_delete() in three route-map community delete handlers (route_set_lcommunity_delete, route_set_community_delete, route_set_ecommunity_delete).

• The intermediate merge variable and the *_uniq_sort() call are eliminated; the dup returned by *_list_match_delete() is used directly as new. This avoids an extra heap allocation and a full sort on every route-map application.
• The claim is valid: each *_list_match_delete() function iterates in forward order, records indices to remove, then deletes them in reverse-index order using memmove — preserving the relative order of survivors. Because the input community is already interned/sorted with no duplicates, the output needs neither a re-sort nor a dedup.

Confidence Score: 5/5

This is a safe, well-reasoned cleanup that removes redundant memory allocations and sorting in three symmetric code paths without changing observable behavior.

The *_list_match_delete functions iterate in forward order and delete entries in reverse-index order via memmove, which provably preserves the sorted order of surviving elements. Because communities are already interned without duplicates, the removed *_uniq_sort call was doing no useful work. Memory ownership transfers cleanly — the dup is now used directly as new with the same lifetime and free/intern path as before. No behavioral difference is introduced.

No files require special attention.

Important Files Changed

Filename	Overview
bgpd/bgp_routemap.c	Removes the intermediate merge variable and the uniq_sort() call in three symmetric community-delete handlers; the deletion functions preserve sorted order in-place, so the optimization is correct.

Sequence Diagram

sequenceDiagram
    participant RM as route_set_*_delete
    participant DUP as community_dup
    participant DEL as *_list_match_delete
    participant SORT as *_uniq_sort (removed)
    participant ATTR as bgp_attr_set_*

    Note over RM,ATTR: Before this PR
    RM->>DUP: dup(old)
    DUP-->>RM: merge (heap alloc)
    RM->>DEL: match_delete(merge, list)
    DEL-->>RM: merge (in-place, sorted)
    RM->>SORT: uniq_sort(merge)
    SORT-->>RM: new (another heap alloc)
    RM->>RM: free(merge)
    RM->>ATTR: set_attr(new) or free(new)

    Note over RM,ATTR: After this PR
    RM->>DUP: dup(old)
    DUP-->>RM: new (heap alloc)
    RM->>DEL: match_delete(new, list)
    DEL-->>RM: new (in-place, sorted preserved)
    RM->>ATTR: set_attr(new) or free(new)

Loading

_{Reviews (1): Last reviewed commit: "bgpd: remove unneeded sort of communitie..." | Re-trigger Greptile}

[mjstapp]

@Mergifyio backport stable/10.6 stable/10.5

[mergify]

backport stable/10.6 stable/10.5

✅ Backports have been created

Details

• #22154 bgpd: remove unneeded sort of communities in rmap delete (backport #22151) has been created for branch stable/10.6
• #22155 bgpd: remove unneeded sort of communities in rmap delete (backport #22151) has been created for branch stable/10.5

[Jafaral]

Thanks!

[riw777]

looks good

[riw777]

It seems like this should be labeled a bug fix?