public/datatransfer: file downloads are corrupt (varying sizes)

[motmot80]

System:

• OS: Windows 11 25H2
• Browser: Firefox 152.0.5 (64-Bit), Chrome 150.0.7871.115 (64-Bit)

Error description

Downloaded zip archives (large) are corrupt and have different sizes every time the attachment is downloaded.

Reproduction path

1 - Check box

Download single zip encrypted archive with

• select an attachment
• download selected files (de: "Ausgewählte Dateien herunterladen")

2 - Download button

Download single zip encrypted archive with

• click on filename
• select "download"

Expected

Downloading a large archive file results in a valid file.

Actual

1100180668 Jul 14 13:55 my-archive.zip
1101047526 Jul 14 13:46 my-archive_2.zip
1100566775 Jul 14 12:57 my-archive_3.zip
1100566716 Jul 14 12:55 my-archive_4.zip

[zhenghongsenqq]

这是来自QQ邮箱的假期自动回复邮件。   您好，我最近正在休假中，无法亲自回复您的邮件。我将在假期结束后，尽快给您回复。

[kreinhard]

Root cause identified and fixed.

The issue was in OakStorage.getFileInputStream() — after obtaining the JCR Binary and returning its InputStream, the method immediately called binary.dispose() in a finally block. This released the underlying storage resource while the stream was still being read by the HTTP response.

For small files this went unnoticed (data already buffered in memory), but for large files the backing data was freed before the download completed, resulting in corrupt files with varying sizes depending on how much data was read before disposal.

Fix: Removed the premature binary.dispose() call. The Binary lifecycle is tied to the JCR Session, which is already properly managed by SessionBoundInputStream

• it keeps the session alive until the download stream is closed.

See commit: 54369643c

[kreinhard]

Thank you for reporting this issue. Open source projects thrive on the support of many contributors.