Conversation
Cyan4973
force-pushed
the
fix3169
branch
2 times, most recently
from
2264e90 to
53aadc7
Compare
Streaming decompression used to wait for a minimum of 5 bytes before attempting decoding. This meant that, in the case that only a few bytes (<5) were provided, and assuming these bytes are incorrect, there would be no error reported. The streaming API would simply request more data, waiting for at least 5 bytes. This PR makes it possible to detect incorrect Frame IDs as soon as the first byte is provided. Fix #3169
terrelln
approved these changes
Jun 22, 2022
Merged
neo-technology-build-agent
pushed a commit
to neo4j/neo4j
that referenced
this pull request
Nov 21, 2025
…) tldr; `Dumper` does not truncate archives when run with `--overwrite-destination`, and Zstd throws you an 'unknown frame descriptor' error for it. In facebook/zstd#3175, released in Zstd 1.5.4, a new error was introduced when the stream decoder has finished decoding a frame. If there are more bytes in the stream, then it assumes that this must correspond to a new frame - if not, then it determines that the archive is corrupt with `Unknown frame descriptor`. Before this error was added, Zstd would quietly ignore that the stream did not contain new necessary marker, `ZSTD_MAGICNUMBER`, or one of the 16 `ZSTD_MAGIC_SKIPPABLE` headers, happily keep consuming the contents. When loading dumps, this can happen as the `TarArchiveInputStream` reaches the `EOF` marker in the TAR archive (two 512 blocks of zeroes). The stream will then attempt to skip to the end of the file.As it does so, it pulls more data from the `ZstdInputStream`, which will attempt to decode a new frame on the trash data. In a correct archive, there should not be more data after the EOF marker, but due to an unfortunate bug, our archives can contain more data under a special circumstance, viz. when running the `Dumper` with the `--override-destination` flag. The root cause is found in the `DefaultFileSystemAbstraction` where the `openAsOutputStream`-method does not truncate files when they are opened with `append=false`, despite its documentation claiming that is should do so. Instead it just opens them for write with file offset 0, meaning that if the original dump was larger than its successor, the successsor will have trash data trailing its valid Zstd frame. As far as I can tell, this issue has been present in Neo4j since 5.19. It seems that the issue was introduced in the refactoring done in neo-technology/neo4j#23520, which changed the `Files.newOutputStream` (which does truncate) to `fs.openAsOutputStream` (which does not). Cherry-picks:
neo-technology-build-agent
pushed a commit
to neo4j/neo4j
that referenced
this pull request
Dec 19, 2025
tldr; `Dumper` does not truncate archives when run with `--overwrite-destination`, and Zstd throws you an 'unknown frame descriptor' error for it. In facebook/zstd#3175, released in Zstd 1.5.4, a new error was introduced when the stream decoder has finished decoding a frame. If there are more bytes in the stream, then it assumes that this must correspond to a new frame - if not, then it determines that the archive is corrupt with `Unknown frame descriptor`. Before this error was added, Zstd would quietly ignore that the stream did not contain new necessary marker, `ZSTD_MAGICNUMBER`, or one of the 16 `ZSTD_MAGIC_SKIPPABLE` headers, happily keep consuming the contents. When loading dumps, this can happen as the `TarArchiveInputStream` reaches the `EOF` marker in the TAR archive (two 512 blocks of zeroes). The stream will then attempt to skip to the end of the file.As it does so, it pulls more data from the `ZstdInputStream`, which will attempt to decode a new frame on the trash data. In a correct archive, there should not be more data after the EOF marker, but due to an unfortunate bug, our archives can contain more data under a special circumstance, viz. when running the `Dumper` with the `--override-destination` flag. The root cause is found in the `DefaultFileSystemAbstraction` where the `openAsOutputStream`-method does not truncate files when they are opened with `append=false`, despite its documentation claiming that is should do so. Instead it just opens them for write with file offset 0, meaning that if the original dump was larger than its successor, the successsor will have trash data trailing its valid Zstd frame. As far as I can tell, this issue has been present in Neo4j since 5.19. It seems that the issue was introduced in the refactoring done in neo-technology/neo4j#23520, which changed the `Files.newOutputStream` (which does truncate) to `fs.openAsOutputStream` (which does not).
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
3 participants
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.
Streaming decompression used to wait for a minimum of 5 bytes before attempting decoding.
This meant that, in the case that only a few bytes (<5) were provided,
and assuming these bytes are incorrect,
there would be no error reported.
The streaming API would simply request more data, waiting for at least 5 bytes.
This PR makes it possible to detect incorrect Frame IDs as soon as the first byte is provided.
Fix #3169 for
urllib3use case