disable frankenphp-trueasync and laravel: TCP fragmentation failures - #1295
Merged
Conversation
Contributor
|
👋 Heads up! This PR modifies the following frameworks:
|
Both fail the fragmentation checks against main, independently of anything
unmerged:
frankenphp-trueasync fails four request shapes -- randomized query,
chunked, chunked in two chunks, randomized body
laravel fails POST chunked and chunked in two chunks
A server that cannot reassemble a request split across TCP segments is not
answering the profile it publishes numbers for.
This PR previously also disabled zix-http2, warp, http4s and fastpysgi-asgi.
That was wrong on all four:
zix-http2, warp fail only the TLS checks added in #1292, which is not
merged. The findings are real -- zix-http2 serves a
self-signed Ed25519 certificate instead of the mounted
RSA pair, warp completes TLS 1.0 handshakes -- but the
rule that forbids them is not in force yet, so disabling
them now punishes entries under a rule that does not
exist. Both pass on main today.
http4s failed POST chunked once on the sweep and passes on
main. One flaky observation is not a verdict.
fastpysgi-asgi fails static and only static on main, so it belongs
with the drop-the-profile treatment in #1294, not here.
MDA2AV
force-pushed
the
chore/disable-nonstatic-failures
branch
from
August 24, 2026 13:53
0963a3d to
349a51c
Compare
MDA2AV
changed the title
MDA2AV
added a commit
that referenced
this pull request
Aug 24, 2026
…1294) * drop the static profiles from the four entries that fail only static From the sweep on #1292. These four fail validation on the static staleness probe and on nothing else, so the entries stay enabled and keep every other profile they publish; they lose static, static-tls and the 21 result keys those produced. fastpysgi-wsgi, uvicorn engine h2o, varnish infrastructure Same treatment #1290 gave the flagship, emerging and experimental entries in this position, applied to the tiers that were left out of it -- with one difference worth stating: for infrastructure this is a real result rather than a formality. h2o and varnish are proxies whose rule explicitly allows open_file_cache and mmap, and they were given a 30s window against the framework tiers' 2s. They still serve the old bytes after it. * also drop the static profiles from fastpysgi-asgi It fails static and only static against main -- 53 passed, 4 failed, every failure a staleness probe -- so it belongs here rather than in #1295, where it was first put. The sweep that classified it ran on the #1292 branch and showed a fragmentation failure that main does not reproduce.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Two entries fail the TCP fragmentation checks against main, independently of anything unmerged. Disabled, 36 result keys removed.
frankenphp-trueasynclaravelA server that cannot reassemble a request split across TCP segments is not answering the profile it publishes numbers for.
Correction — this PR was wrong before
It originally disabled four more entries. All four were mistakes, and the CI run on this branch is what showed it:
zix-http2andwarpfail only the TLS checks added in validate: check TLS quality and TLS fairness, not just ALPN #1292, which is not merged. The findings are real — zix-http2 serves a self-signed Ed25519 certificate instead of the mounted RSA pair (Dockerfile:64), and warp completes TLS 1.0 handshakes — but the rule forbidding them is not in force yet. Disabling them now punishes entries under a rule that does not exist. Both pass on main today. These belong with validate: check TLS quality and TLS fairness, not just ALPN #1292, once it merges.http4sfailed POST chunked once on the sweep and passes on main. One flaky observation is not a verdict.fastpysgi-asgifails static and only static on main (53 passed / 4 failed, every failure a staleness probe), so it belongs with the drop-the-profile treatment in drop the static profiles from the four entries that fail only static #1294 rather than being disabled.The original classification came from the sweep run on the #1292 branch, which carries checks main does not have. Sorting failures by that sweep without re-checking them against main is what produced the error.