Test file generator

An upload limit is a boundary, so it needs boundary values: a file exactly at the limit, one byte under and one byte over. A limit of 10 MB is tested with files of 9,999,999, 10,000,000 and 10,000,001 bytes, and because some systems mean 10 MiB, also with 10,485,760 and 10,485,761 bytes. A "roughly 10 MB" file from a downloads folder cannot tell you which side of the limit the system really enforces.

Every file here is built to the byte. PDFs are padded with an unused object, PNG and JPEG images with a metadata block that viewers ignore, and CSV and JSON files by filling the last row, so each file is still valid for its format at the exact size you asked for.

A megabyte means 1,000,000 bytes in the SI sense and 1,048,576 bytes in the binary sense, which is properly called a mebibyte (MiB). Operating systems, browsers, cloud storage and upload libraries disagree on which they mean, and the gap is almost 5% at every step. A file of 10,200,000 bytes is over a 10 MB limit and under a 10 MiB one.

Choose the unit the requirement uses, then test the other one as well. Mismatches between the front end, which often checks in one unit, and the server, which checks in another, are a common source of uploads that pass validation and then fail.

Most upload defects come from files that are valid in an unexpected way. An empty file, an image saved with a .pdf extension, a PDF renamed to photo.jpg, a file cut off halfway, a name with no extension, an upper-case extension, a name with emoji or 255 characters: each tests an assumption the code makes about what a file looks like.

The edge cases also cover security checks. A double extension such as report.pdf.exe confirms the filter reads the real extension, and the pixel flood image, a small PNG that decodes to 400 million pixels, confirms the image pipeline limits dimensions before it decodes.

A file of zeros or repeated text compresses to almost nothing. If a proxy, CDN or the browser compresses the upload, a 100 MB file of zeros can arrive as a few hundred kilobytes and never reach the limit you meant to test. Random bytes cannot be compressed, so they test the limit as it is enforced on the wire.

Use random contents for size and timeout tests, and zeros or a text pattern when you want a file that compresses well, for example to check that compression is actually applied.