@@ -4,18 +4,18 @@ internal static class TarArchiveDefaults
44 {
55 // A block factor of 1 keeps the record size equal to the block size (512 B),
66 // so SharpZipLib writes no extra zero padding beyond the two standard EOF
7- // blocks. The default factor of 20 produces ~8 KB of trailing zeros, which
8- // can trigger a race in Podman's archive handler. The tar subprocess exits
9- // after the EOF blocks while the HTTP sender is still flushing the padding,
10- // resulting in an EPIPE (HTTP 500 "broken pipe"). See:
11- // https://github.com/testcontainers/testcontainers-dotnet/issues/1683.
7+ // blocks. The default factor of 20 produces up to 9 KiB of trailing zero
8+ // padding, which can trigger a race in Podman's archive handler. The tar
9+ // subprocess exits after the EOF blocks while the HTTP sender is still
10+ // flushing the padding, resulting in an EPIPE (HTTP 500 "broken pipe").
11+ // See: https://github.com/testcontainers/testcontainers-dotnet/issues/1683.
1212 //
1313 // The race only affects PUT /containers/{id}/archive (resource mappings via
1414 // a MemoryStream). Buildah's copierHandlerPut returned immediately after
1515 // tar.Reader reached io.EOF without draining the pipe. POST /build is immune
1616 // because containers/storage's chrootarchive.Untar always drains stdin after
17- // extraction. The server-side fix landed in buildah. See:
18- // https://github.com/podman-container-tools/buildah/pull/6678.
17+ // extraction. The server-side fix landed in buildah.
18+ // See: https://github.com/podman-container-tools/buildah/pull/6678.
1919 // Keeping the block factor at 1 stays correct for older Podman versions.
2020 //
2121 // Do NOT use this for the Dockerfile build context, which is written to a
0 commit comments