forcing color from Maven side is not an option: the option is to fix JAnsi
Did you open an issue or even a PR against JAnsi?
We'll integrate fixed JAnsi release.

Notice that for future Maven 3.5.1, we already integrated a new JAnsi release that should fix some issues: did you check if Maven 3.5.1-SNAPSHOT is ok in your conditions?

Please note, this work around seems effective, at least for me:

https://plus.google.com/111012722058314768782/posts/Z8ufa5LS6nv

Perhaps while waiting on JAnsi change for Cygwin, the mvn script could be updated to include adding `-Djansi.passthrough=true` to MAVEN_OPTS?

thanks for your positive report
On adding the option to mvn script, I fear this is not an option: this would affect any Unix user (Linux, Mac, FreeBSD, ...) and not only Windows+Cygwin users
Notice that, after releasing Maven 3.5.1 where we have MNG-6282 that is a wider regression, if we release Maven 3.5.2 without Jansi fix, I'll add in Maven the workaround code I proposed in https://github.com/fusesource/jansi/issues/94#issuecomment-329977920
Then whatever happens, Maven will have an improved Jansi detection, be it done at any level

Ah, just so. Were the passthrough flag to be added, it would be with a uname check for Cygwin.

To be specific:

*$ diff mvn-original mvn
183a184,189
> if $cygwin || $mingw; then
>   if [ -t 1 ]; then
>     MAVEN_OPTS="-Djansi.passthrough=true $MAVEN_OPTS"
>   fi
> fi
>

ok, I:

• didn't remember these $cygwin and $mingw variables that were available in mvn shell script
• didn't think about the test -t DESCRIPTOR to detect terminal

ok, if I need to code a workaround for Maven 3.5.2, I buy your implementation which is perfect
thanks for the detailed hint