I made a fresh clone from master today, and looked into it myself.
In CommandLineHelpTest.java all the tests that is failing is instanciating
a Help class. In all cases it have an alternative constructor where you can
specify Help.Ansi.OFF to avoid that the test depends on the console it run
from.
CommandLineTest did also have 3 faults, and here it is CommandLine.run()
that needs Help.Ansi.OFF as it's third argument. I can send you patch if
you tell me how you want it.
Atle
Den søn. 11. mar. 2018 kl. 11:04 skrev Remko Popma <remko.po...@gmail.com>:
> Thanks for reporting this. I’ll take a look.
>
> Remko
>
> > On Mar 11, 2018, at 17:32, Atle Tokle <ato...@gmail.com> wrote:
> >
> > I tried to build from GIT bash on windows 10, and first test that failed
> > was
> >
> > testSynopsisOrderCorrectWhenParametersDeclaredOutOfOrder in
> >
> > org.apache.logging.log4j.core.tools.picocli.
> >
> > CommandLineHelpTest:2008
> >
> >
> > It seems like the expected result
> >
> > format("<main class> <param0> <param1>%n") on line 2014 always assumes
> > that no ANSI color codes is output, while the actual results have ANSI
> > codes.
> >
> > I looked into the code, and saw that one of the things that was check
> > was if TERM environment variable started with xterm, and did a test
> > where I first did
> >
> > export TERM=zzzzz
> >
> > and did then build it again. Now without faults.
> >
> > The test should have more control on what environment it is running
> > on. It is several other tests in the same testclass that fails because
> > of same issue.
> >
> >
> > I guess this is the same fault that Gary Gregory experienced here
> > https://www.mail-archive.com/dev@logging.apache.org/msg05223.html
> >
> > in message from Nov 19, 2017, at 6:32 PM
> >
> >
> > Best regards
> >
> > Atle
>
