fix: Make sure the primary path exists at various steps in the handshake

[larseggert]

Fixes #1448

[github-actions]

Benchmark results

Performance differences relative to c004359.

• drain a timer quickly time: [429.31 ns 435.00 ns 440.29 ns]
  change: [+0.5498% +1.7720% +3.0422%] (p = 0.01 < 0.05)
  Change within noise threshold.

• coalesce_acked_from_zero 1+1 entries
  time: [194.83 ns 195.49 ns 196.35 ns]
  change: [-3.2560% -1.5610% +1.2686%] (p = 0.28 > 0.05)
  No change in performance detected.

• coalesce_acked_from_zero 3+1 entries
  time: [239.31 ns 239.87 ns 240.52 ns]
  change: [-0.8737% -0.6225% -0.3609%] (p = 0.00 < 0.05)
  Change within noise threshold.

• coalesce_acked_from_zero 10+1 entries
  time: [238.44 ns 239.71 ns 241.11 ns]
  change: [-1.4186% -0.9377% -0.4908%] (p = 0.00 < 0.05)
  Change within noise threshold.

• coalesce_acked_from_zero 1000+1 entries
  time: [219.88 ns 220.04 ns 220.25 ns]
  change: [-13.922% -5.0020% +0.4701%] (p = 0.40 > 0.05)
  No change in performance detected.

• RxStreamOrderer::inbound_frame()
  time: [118.62 ms 118.76 ms 118.99 ms]
  change: [-0.1152% +0.0491% +0.2646%] (p = 0.65 > 0.05)
  No change in performance detected.

• transfer/Run multiple transfers with varying seeds
  time: [118.92 ms 119.23 ms 119.55 ms]
  thrpt: [33.458 MiB/s 33.548 MiB/s 33.635 MiB/s]
  change:
  time: [-0.9682% -0.6168% -0.2464%] (p = 0.00 < 0.05)
  thrpt: [+0.2470% +0.6206% +0.9776%]
  Change within noise threshold.

• transfer/Run multiple transfers with the same seed
  time: [119.92 ms 120.12 ms 120.32 ms]
  thrpt: [33.244 MiB/s 33.300 MiB/s 33.354 MiB/s]
  change:
  time: [-0.7662% -0.5401% -0.3157%] (p = 0.00 < 0.05)
  thrpt: [+0.3167% +0.5431% +0.7721%]
  Change within noise threshold.

• 1-conn/1-100mb-resp (aka. Download)/client
  time: [1.1213 s 1.1276 s 1.1340 s]
  thrpt: [88.187 MiB/s 88.683 MiB/s 89.183 MiB/s]
  change:
  time: [+3.2833% +4.2519% +5.1445%] (p = 0.00 < 0.05)
  thrpt: [-4.8928% -4.0785% -3.1790%]
  💔 Performance has regressed.

• 1-conn/10_000-parallel-1b-resp (aka. RPS)/client
  time: [427.10 ms 429.27 ms 431.43 ms]
  thrpt: [23.179 Kelem/s 23.296 Kelem/s 23.414 Kelem/s]
  change:
  time: [-0.4027% +0.2569% +0.9424%] (p = 0.47 > 0.05)
  thrpt: [-0.9336% -0.2562% +0.4043%]
  No change in performance detected.

• 1-conn/1-1b-resp (aka. HPS)/client
  time: [51.754 ms 51.956 ms 52.213 ms]
  thrpt: [19.152 elem/s 19.247 elem/s 19.322 elem/s]
  change:
  time: [-1.6277% -0.7624% +0.1402%] (p = 0.09 > 0.05)
  thrpt: [-0.1400% +0.7683% +1.6547%]
  No change in performance detected.

Client/server transfer results

Transfer of 134217728 bytes over loopback.

Client	Server	CC	Pacing	Mean [ms]	Min [ms]	Max [ms]	Relative
msquic	msquic			804.0 ± 372.0	385.4	1252.5	1.00
neqo	msquic	reno	on	931.5 ± 255.4	770.8	1426.3	1.00
neqo	msquic	reno		941.6 ± 161.5	764.1	1181.6	1.00
neqo	msquic	cubic	on	979.8 ± 246.2	776.0	1417.9	1.00
neqo	msquic	cubic		1037.1 ± 205.0	812.4	1428.7	1.00
msquic	neqo	reno	on	4448.2 ± 218.2	4105.7	4766.9	1.00
msquic	neqo	reno		4384.5 ± 256.7	4036.4	4891.2	1.00
msquic	neqo	cubic	on	4456.9 ± 237.2	4190.7	4999.2	1.00
msquic	neqo	cubic		4451.3 ± 222.1	4159.4	4896.2	1.00
neqo	neqo	reno	on	3796.4 ± 277.7	3427.2	4446.3	1.00
neqo	neqo	reno		3554.2 ± 157.9	3274.6	3789.5	1.00
neqo	neqo	cubic	on	4250.1 ± 594.8	2875.4	5305.3	1.00
neqo	neqo	cubic		4378.6 ± 215.3	4095.2	4835.6	1.00

⬇️ Download logs

[codecov]

Codecov Report

Attention: Patch coverage is 91.66667% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 93.13%. Comparing base (c44e536) to head (9408049).
Report is 4 commits behind head on main.

Files	Patch %	Lines
neqo-transport/src/connection/mod.rs	91.66%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1814   +/-   ##
=======================================
  Coverage   93.13%   93.13%           
=======================================
  Files         117      117           
  Lines       36339    36354   +15     
=======================================
+ Hits        33843    33858   +15     
  Misses       2496     2496           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[martinthomson]

The coverage complaint seems like a reasonable trade.