Durable runtime response-length fix — kill the truncation bug CLASS (generalize #79) #85

New Issue

2026-08-03T18:55:14Z

will.anderson commented

2026-08-03 18:55:14 +00:00

Tag: POST-BETA · ORTHOGONAL · important
Owner: @tim.lingo

Why

#79 fixed one truncation via a pointer-identity fix, but the underlying class remains: any handler that fs_reads and then returns a body longer than what it read can truncate the reply. This signature has recurred (neuron #96 safety-contact, neuron-ui #184 Windows stale runtime). Per-handler workarounds don't close the class.

Scope

Generalize the fix in the el runtime so response length is governed by the actual body, not by an earlier read buffer / pointer:

Audit the HTTP response path for any place a length is carried from a prior fs_read/buffer rather than recomputed from the emitted body.
Make the durable, runtime-level fix so no handler can truncate this way, regardless of how it was written.

Acceptance criteria

A regression test: a handler that reads a short file then returns a long body returns the full body.
The fix is in the runtime response path, not in individual handlers.
Reference: builds on #79's pointer-identity fix.

**Tag:** POST-BETA · ORTHOGONAL · **important** **Owner:** @tim.lingo ## Why #79 fixed one truncation via a pointer-identity fix, but the underlying **class** remains: **any handler that `fs_read`s and then returns a body longer than what it read can truncate** the reply. This signature has recurred (neuron #96 safety-contact, neuron-ui #184 Windows stale runtime). Per-handler workarounds don't close the class. ## Scope Generalize the fix in the **el runtime** so response length is governed by the actual body, not by an earlier read buffer / pointer: - Audit the HTTP response path for any place a length is carried from a prior `fs_read`/buffer rather than recomputed from the emitted body. - Make the durable, runtime-level fix so **no** handler can truncate this way, regardless of how it was written. ## Acceptance criteria - A regression test: a handler that reads a short file then returns a long body returns the **full** body. - The fix is in the runtime response path, not in individual handlers. - Reference: builds on #79's pointer-identity fix.

will.anderson added the POST-BETA ORTHOGONAL labels 2026-08-03 18:55:14 +00:00

tim.lingo was assigned by will.anderson

2026-08-03 18:55:14 +00:00

will.anderson referenced this issue

2026-08-03 19:00:02 +00:00

soul RSS leak ~35MB/min: awareness-loop allocations bypass the per-request arena (OOM-cycled hourly; likely hits GKE too) #70

will.anderson commented

2026-08-03 19:17:02 +00:00

@tim.lingo Confirm the scope with you: the durable fix belongs in the el runtime response path (length governed by the emitted body, not a prior fs_read buffer), not per-handler — agreed? Please coordinate with #79 (the pointer-identity fix this builds on) and #70 (the RSS leak, separate bug).

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: neuron-technologies/el#85