Re: Checking in files with trailing white space

[Paul Eggert]

On 12/14/2016 07:32 AM, Eli Zaretskii wrote:
> > Cc: address@hidden, address@hidden, address@hidden
> > From: Paul Eggert <address@hidden>
> > Date: Tue, 13 Dec 2016 23:47:02 -0800
> >
> > Eli Zaretskii wrote:
> > > providing compressed text files in the distribution
> > > doesn't violate the GPL or its spirit.
> > Sure, but we are talking about binary data files that are in the repository,
> > which is not the same thing. Most of these files are not in .gz form; they are
> > typically object files (in the GPL sense) of some other sort.
> Which files are those?

I've attached a list of repository-controlled files that are not UTF-8 
text files in the POSIX sense. Of the 433 such files, I count 181 Netpbm 
image data, 107 gzipped files (mostly gzipped text), 57 PNG files, 11 
PDF files, and there is a hodgepodge of assorted other formats. A few 
files are in the list merely because they lack a trailing newline (a 
problem for some tools, and POSIX says such files are not text files). 
(For a sense of the scale here, all this is out of 3695 files total in 
Emacs master right now.)

Most of the files in this list are merely awkward and are not a GPL 
issue; these include the gzipped text files, the text files that use 
non-UTF-8 encodings, and the no-trailing-newline files. And many files 
are there because we have put object files into the source repository, 
where we really should be generate them as part of the build and limit 
the repository to source code; these include the PBM and PNG files (or 
at least I hope they do; I haven't checked).

> Depends on the compression method, I guess. For some of them, we
> cannot rely on the respective tools being available, so if that
> particular kind of compression is needed for some test, having the
> source file might not be good enough.

We can assume gzip; it's universally available and is needed anyway to 
build tarballs. This will handle all the compressed files.

not-UTF-8-txt.ls
Description: Text document