On 25 May 2016 10:54 p.m., "Gary Gregory" <garydgreg...@gmail.com> wrote: > > 0 svn only matches src if I ignore whitespace (CRLF added in src > > archives) -- also src/test/**/CSVBenchmark.java is missing from src - > > why? > CR vs. CRLF is important for the test data since the parser and writer can > be configured to use specific line endings. I can configured a writer to > output with a specific char or string as the record-ending string.
OK, perhaps for this test data it is important to have specific CRLF and LF (perhaps even a nasty test case with both - I'll tell you in CSV world we find many weird things! :) Different character encoding per column anyone?) In that case, should set the eol property explicitly for each *csv test file in svn? Or perhaps just for explicit CRLF test files. As the current tests pass in either CRLF and LF case this does not hold up the release. :-) It still surprised me that the release plugin checked out with CRLF for all files in the dist archive, while my svn co gives LF - but I might have messed up my local svn setup.