Re: Current trunk fails to build on MS-Windows

[Eli Zaretskii]

> From: address@hidden
> Date: Sun, 11 Jan 2015 03:11:22 -0700
> Cc: address@hidden
> 
> > (Why is it necessary to go through a temporary file, anyway?)
> 
> jrtangle / jrweave are a literate programming system. In the general
> case, multiple files may be generated, and if only one file has changed,
> only that file should be updated, so that make doesn't recompile
> the world.

I understand that.  But I see that in dump_files you already have in
memory the contents of the file you are about to write.  Why not
compare it with the current file on disk (thus emulating "cmp -s"),
and refrain from rewriting the file if the comparison succeeds?  If
the comparison fails, you can then rewrite without fear.  Am I missing
something?

> That said, there is provision for replacing the default recipe
> that uses a temp file to compare with what's on disk; for this case
> simply always generating the final output will be fine.
> 
> I will try to get to this during the course of this week.
> 
> Thanks for pointing out the issue. It's worth fixing.

Thanks in advance.

As "plan B", I already have a variant of jrtangle that includes a
special recipe for MS-Windows, which does the same as on Posix hosts:
compare the old and the new files, and if different, move the new over
the old, using native Windows facilities.  If none of the above
options work out, we can use this fallback, I think.

> P.S. If you're interested, see http//github.com/arnoldrobins/texiwebjr
> for the literate source for jrweave and jrtangle.  The PDF that gets
> generated is a combination literate program and user's manual.

Thanks, I already have that checked-out.