The test "re-encodes smaller until the thumbnail fits the recorded size" built a noisy 400x300 JPEG and encoded it several times, which came close to vitest's 5 s limit and timed out on a busy host. It now uses a noisy 64x48 JPEG.
The proof is unchanged. The image is below every encoding's maximum dimension, so no encoding resizes it, and the recorded size is one byte below the ciphertext of the first, default encoding (quality 50), so that encoding cannot fit. The test still asserts that a smaller thumbnail that fits was uploaded. No timeout, skip or retry was added, and every assertion is kept.
Measured test time: 0.9 to 1.2 s per build, 2.3 s at host load about 160; it was 5.4 to 6.1 s before.
This is the next half of #153; main gets the same change in its own PR.
Model: opus-5-5
The test "re-encodes smaller until the thumbnail fits the recorded size" built a noisy 400x300 JPEG and encoded it several times, which came close to vitest's 5 s limit and timed out on a busy host. It now uses a noisy 64x48 JPEG.
The proof is unchanged. The image is below every encoding's maximum dimension, so no encoding resizes it, and the recorded size is one byte below the ciphertext of the first, default encoding (quality 50), so that encoding cannot fit. The test still asserts that a smaller thumbnail that fits was uploaded. No timeout, skip or retry was added, and every assertion is kept.
Measured test time: 0.9 to 1.2 s per build, 2.3 s at host load about 160; it was 5.4 to 6.1 s before.
This is the `next` half of https://git.eeqj.de/sneak/quak/issues/153; `main` gets the same change in its own PR.
Model: opus-5-5
The test "re-encodes smaller until the thumbnail fits the recorded size"
built a noisy 400x300 JPEG and encoded it several times, which took close
to the 5 s test limit and timed out on a busy host. It now uses a noisy
64x48 JPEG. The recorded size is still one byte less than the default
quality-50 encoding, so the first encoding still cannot fit and the test
still proves that quak re-encodes smaller until it fits. Every assertion
is unchanged.
Model: opus-5-5
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
The test "re-encodes smaller until the thumbnail fits the recorded size" built a noisy 400x300 JPEG and encoded it several times, which came close to vitest's 5 s limit and timed out on a busy host. It now uses a noisy 64x48 JPEG.
The proof is unchanged. The image is below every encoding's maximum dimension, so no encoding resizes it, and the recorded size is one byte below the ciphertext of the first, default encoding (quality 50), so that encoding cannot fit. The test still asserts that a smaller thumbnail that fits was uploaded. No timeout, skip or retry was added, and every assertion is kept.
Measured test time: 0.9 to 1.2 s per build, 2.3 s at host load about 160; it was 5.4 to 6.1 s before.
This is the
nexthalf of #153;maingets the same change in its own PR.Model: opus-5-5
Review: pass.
Model: opus-5-5