echo icon indicating copy to clipboard operation
echo copied to clipboard

static middleware: path ist unescaped twice for file names, leading to not downloadable content

Open georgmu opened this issue 1 year ago • 2 comments

Issue Description

Given a file on disk with a percent sign in its name, it is not possible to download it using echo's static middleware.

Given some file names:

  • 100%.txt
  • foo%20bar.txt (this is really the name on disk)

It is not possible to download these files using the static middleware.

Checklist

  • [X] Dependencies installed
  • [X] No typos
  • [X] Searched existing issues and docs

Expected behaviour

The standard way to download them would be using the folowing paths:

  • 100%.txt -> GET /100%25.txt
  • foo%20bar.txt -> GET /foo%2520bar.txt

Actual behaviour

  • 100%.txt -> GET /100%25.txt
    • echo error message: invalid URL escape "%.p (because of the double unescape and .p is no valid hex)
  • foo%20bar.txt -> GET /foo%2520bar.txt
    • echo error message: path does not exist (because file foo bar.txt does not exist, only foo%20bar.txt)

Background is that url.Path in http.Request is already escaped (as the documentation for url.Path suggests), but echo is unescaping it again.

I tried to generate a fix. For the non-embedded case this is solved by simply removing the explicit url.PathUnescape call in static.go (see https://github.com/georgmu/echo/commit/852dedeca2a22a0b6967e71f69cbb0823959a198 which also contains some test cases). I haven't created a merge request yet, as the embedding using groups or path patterns is a bit more complicated (I haven't fully understand the case for the disablePathUnescaping toggle).

Steps to reproduce / Working code to debug

See test cases in https://github.com/georgmu/echo/commit/852dedeca2a22a0b6967e71f69cbb0823959a198

Version/commit

master ( commit fa70db8 )

georgmu avatar Feb 29 '24 11:02 georgmu