[
https://issues.apache.org/jira/browse/IVY-1522?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16018590#comment-16018590
]
ASF GitHub Bot commented on IVY-1522:
-------------------------------------
Github user twogee commented on the issue:
https://github.com/apache/ant-ivy/pull/20
I can assure you that colon in version has NEVER have worked on Windows.
Test case
```
import java.io.*;
public class TestColonNTFS {
public static void main(String[] args) throws IOException {
try (Writer writer = new BufferedWriter(new OutputStreamWriter(new
FileOutputStream("file:name.txt"), "utf-8"))) {
writer.write("something");
}
}
}
```
produces
```
>dir
...
2017-05-20 20:39 0 file
...
>dir /r file
...
2017-05-20 20:39 0 file
9 file:name.txt:$DATA
```
And that is the correct behaviour of NTFS: create an [alternate data
stream](https://msdn.microsoft.com/en-us/library/windows/desktop/aa364404). You
can use that for versioning, for sure ;-) But this is not portable.
Now, should we introduce a special treatment for reserved and unsafe URL
characters, we will create a lock-in unless all other tools accept the same
conventions, which is unlikely. And that defeats the very idea of flexibility.
> Ivy 2.4.0 improperly handles modules with colon (:) in version
> --------------------------------------------------------------
>
> Key: IVY-1522
> URL: https://issues.apache.org/jira/browse/IVY-1522
> Project: Ivy
> Issue Type: Bug
> Components: Core
> Affects Versions: 2.4.0
> Reporter: Daniel Holmes
> Priority: Critical
>
> This was not an issue in 2.2.0.
> I have some modules in my project which produce debian .deb files and those
> modules versions include a colon ':' in the version because the version
> scheme was reset, so an epoch is set.
> Also for this module the ivy.xml declaration uses extends tag, and in
> processing the parent location, this code in FileUtil which is called during
> a FileUtil.normalize() is causing the : in the resolve parent file name to
> have the ':' replaced with a ':/'.
> The comment on the it seems to explain that it is trying to handle a
> Windows/NetWare case, but the assumption to process any colon in the path is
> too broad and is negatively affecting my case which is executing on a linux
> system. Even on windows, this would only be attempting to address a drive
> letter case, right, where colon would occur only after a single letter in the
> path?
> private static String[] dissect(String path) {
> char sep = File.separatorChar;
> path = path.replace('/', sep).replace('\\', sep);
> // // make sure we are dealing with an absolute path
> // if (!isAbsolutePath(path)) {
> // throw new BuildException(path + " is not an absolute path");
> // }
> String root = null;
> int colon = path.indexOf(':');
> if (colon > 0) { // && (ON_DOS || ON_NETWARE)) {
> int next = colon + 1;
> root = path.substring(0, next);
> char[] ca = path.toCharArray();
> root += sep;
> // remove the initial separator; the root has it.
> next = (ca[next] == sep) ? next + 1 : next;
> StringBuffer sbPath = new StringBuffer();
> // Eliminate consecutive slashes after the drive spec:
> for (int i = next; i < ca.length; i++) {
> if (ca[i] != sep || ca[i - 1] != sep) {
> sbPath.append(ca[i]);
> }
> }
> path = sbPath.toString();
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
