Johannes Schindelin <johannes.schindelin@xxxxxx> writes: > Note: the test suite also uses the t/README file as well as the COPYING > file in t/diff-lib/, expecting LF-only line endings explicitly and > failing if that assumption does not hold true. That is why we mark them > as LF-only in the .gitattributes, too. I said the previous step that used COPYING was good because it didn't force the file to be checked out with lf (and instead fixed the test to strip CR if/as necessary), but come to think of it, these COPYING/README files are in t/diff-lib/ that are shipped as test vector, and meant to stay constant even when the end-user facing COPYING and README at the top level changed. I do not see t/diff-lib/* marked as eol=lf in this patch, but shouldn't it be done here, just like all these test vector files? I also wonder if that makes the previous step unnecessary. > > This patch can be validated even on Linux by using this cadence: > > git config core.autocrlf true > rm .git/index && git stash > make -j15 DEVELOPER=1 test > > Reviewed-by: Jonathan Nieder <jrnieder@xxxxxxxxx> > Signed-off-by: Johannes Schindelin <johannes.schindelin@xxxxxx> > --- > t/.gitattributes | 20 +++++++++++++++++++- > 1 file changed, 19 insertions(+), 1 deletion(-) > > diff --git a/t/.gitattributes b/t/.gitattributes > index 2d44088f56e..3525ca43f30 100644 > --- a/t/.gitattributes > +++ b/t/.gitattributes > @@ -1,2 +1,20 @@ > t[0-9][0-9][0-9][0-9]/* -whitespace > -t0110/url-* binary > +/t0110/url-* binary > +/t3900/*.txt eol=lf > +/t3901/*.txt eol=lf > +/t4034/*/* eol=lf > +/t4013/* eol=lf > +/t4018/* eol=lf > +/t4100/* eol=lf > +/t4101/* eol=lf > +/t4109/* eol=lf > +/t4110/* eol=lf > +/t4135/* eol=lf > +/t4211/* eol=lf > +/t4252/* eol=lf > +/t5100/* eol=lf > +/t5515/* eol=lf > +/t556x_common eol=lf > +/t7500/* eol=lf > +/t8005/*.txt eol=lf > +/t9*/*.dump eol=lf