Re: [VFS] Possible regression in path decoding between 2.9.0 and 2.10.0

Hi everyone,

Just following up on this to see if anybody has any insight into this
issue and whether it is a bug or not. Workaround suggestions would
also be appreciated.

Thanks again,
Nick

On Wed, Jun 17, 2026 at 7:57=E2=80=AFPM Nick Tarallo <[email protected]=
m> wrote:
>
> Hello,
>
> I have encountered a possible regression between VFS 2.9.0 and 2.10.0
> related to file path encoding.
>
> I would expect resolving a file from a URL and then resolving the URL
> returned by getURL() to preserve the same path, but that does not
> appear to happen in 2.10.0. The following test case from my project
> passes with 2.9.0 but fails with 2.10.0:
>
> @Test
> public void vfsGetUrlRoundTripIdentity() throws Exception {
>    Path tmp =3D Files.createTempDirectory("vfs-roundtrip");
>    try {
>       Path child =3D tmp.resolve("outside%text[inside%text]tail");
>       Files.createDirectories(child);
>
>       FileSystemManager mgr =3D VFS.getManager();
>       FileObject a =3D mgr.resolveFile(child.toUri().toString());
>       FileObject b =3D mgr.resolveFile(a.getURL().toString());
>
>       assertEquals(a.getName().getPath(), b.getName().getPath());
>    } finally {
>       FileUtils.deleteDirectory(tmp.toFile());
>    }
> }
>
> The assertion failure is:
>
> Expected:
> .../outside%25text[inside%25text]tail
>
> Actual:
> .../outside%25text[inside%2525text]tail
>
> I noticed that URIParser.decode() changed in PR #438 [1], and I am
> wondering whether the behavior could be related to that change. It
> appears that content inside brackets is treated specially to avoid
> decoding IPv6 hosts, but I am not sure whether that logic should apply
> to local file paths.
>
> Is this expected behavior, or does it sound like a regression?
>
> Thanks,
> Nick
>
> [1] https://github.com/apache/commons-vfs/pull/438/changes#diff-731c9fd14=
9c7d934456163a6eb8c1afebc6be0669986e8ccf5092f164ea0cd30R166