Problem with Data Integrity Protections for Amazon S3

[ErwanLeroux]

With version 2.30.0 of https://mvnrepository.com/artifact/software.amazon.awssdk/bom/
byte[] tarContent = s3Resource.getContentAsByteArray(); returns more than the previously uploaded file, resulting in a corrupted file.

I found this issue :

And configured my S3Client with

S3Client.builder()
.responseChecksumValidation(ResponseChecksumValidation.WHEN_REQUIRED)
.requestChecksumCalculation(RequestChecksumCalculation.WHEN_REQUIRED)
.build();

That made byte[] tarContent = s3Resource.getContentAsByteArray(); return the exact file.

It appears to be linked to aws/aws-sdk-java-v2#5802

Could you look into it, to see if there anything to do beside configuration ?

[Robothy]

@ErwanLeroux Thanks for reporting this issue. Please try to resolve it by upgrading the LocalS3 version to 1.20.

[ErwanLeroux]

Sadly, it doesn't work without setting

                .responseChecksumValidation(ResponseChecksumValidation.WHEN_REQUIRED)
                .requestChecksumCalculation(RequestChecksumCalculation.WHEN_REQUIRED)

even with the 1.20 version

[ErwanLeroux]

For security reason, I can't provide with the full file that cause the problem, but I can show you the extraneous bytes at the start of the file
"20000;chunk-signature=c9e25a83116f8f2210860984da2119e9cb0e0e6061898a6b125aa817bd9155ca\nEPE"
I'm not sure if there really is a return char between 'ca' and 'EPE', it could a misinterpretation from my binary visualiser.
In bytes it's '0D 0A'

[Robothy]

@ErwanLeroux It appears that LocalS3 is not decoding the stream correctly when uploading objects. Could you please provide a test code snippet to reproduce this issue?