[CZT-Devel] SpecReader failing on Windows

[Andrius V. <an...@ve...>]

Hi everyone,

A recent query to czt-devel by Jon Lockhart hinted towards a problem
on building CZT on Windows. Previously, CZT build was checked out with
Unix-style line breaks (LF). For better compatibility, on Windows,
Windows-style line breaks (CR+LF) should be used (new files are
usually created with Windows-style breaks). I have updated the Git
repository to handle this: on Windows, CR+LF will be checked-out for
text files, and the repository will always contain only LF.

The problem arises with SpecReader module. It expects single-character
line endings (i.e. LF: "\n" instead of "\r\n" for Windows). Therefore
it calculates some character positions incorrectly and fails. I have
added a quick workaround that allows tests to go through (using
buffered SpecReader), but part of its functionality still remains
erroneous. I have filed this with more details as bug #3545808
(https://sourceforge.net/tracker/?func=detail&aid=3545808&group_id=86250&atid=578942).

If anyone is more familiar with SpecReader functionality, it would be
nice to correct the problems (the original author is Ian Toyn).
Otherwise, the module does not seem to be used much - what are its
dependencies and its role in the core CZT?

All other modules are working on Windows with CR+LF source/test files.

Best regards,
~Andrius