Re: GNU make 3.81beta3 released

[Eli Zaretskii]

> Date: Tue, 28 Jun 2005 22:05:58 -0400
> From: "Paul D. Smith" <address@hidden>
> Cc: address@hidden
> 
> OK, looking at the first failed regression test I happened upon, the
> only difference is that the "base" file, which is what the regression
> suite (perl) prints out as the "known-good output", is using UNIX line
> endings (LF), while the "log" file, which is what GNU make prints out,
> is using DOS/Windows line endings (CRLF).
> 
> 
> This rings a bell with me; I think we've run into this before.

Yes, we have.

> Anyone have any thoughts on how to resolve this issue?  We need
> to get Perl to generate files with CRLF line endings, or GNU make to
> generate output with LF line endings, or diff to ignore lines that are
> different only by having a CR (^M) at the end.

Yes, either of these would work, although I'd expect Diff to ignore
line-end format by default.  Earnie, what port of Diff you used?  For
that matter, what port of Perl?

I don't like the alternative with Make producing Unix-style EOLs: it
will be very tricky because some of the output comes from programs
Make invokes, and we cannot always fix all of those.

Also, there's another solution: run all known good output files
through unix2dos before running the test suite.