Re: texi2dvi: set -e, use tar

[Eli Zaretskii]

> Date: Mon, 9 Jan 2006 16:13:16 -0600
> From: address@hidden (Karl Berry)
> Cc: address@hidden, address@hidden
> 
>     Can you please describe the situation where texi2dvi is run on a
>     directory tree like the comment above suggests?  I didn't think this
>     could happen.
> 
> I believe a LaTeX document might be set up that way.  However, perhaps
> cp -pr would be simpler.

It would, if -p and -r are widely supported.  Are they?

>     The problem with this replacement is that on non-Posix platforms it
>     requires yet another ported utility (Tar) 
> 
> Which is everywhere, right?  I find it quite hard to imagine a system,
> even MS-DOS, which has sufficient support to run texi2dvi (e.g., has
> tex), but lacks tar.

The issue is not whether there is a port of Tar, but whether it is
installed on the end-user's machine.

>     and a very reliable support in the shell for Posix-style
>     (command...) semantics.
> 
> It's a pretty basic shell construct.

That requires a faithful emulation of `fork'.

> Either the shell is implemented or not.  If it's not, I doubt
> texi2dvi will run at all.  Heck, it makes use of tons of
> user-defined functions now, which are a lot more obscure feature
> than | and &&.

It strikes me that we should reimplement texi2dvi as a C program.
Beyond certain level of complexity (which IMHO texi2dvi passed way
back), there's no reasons to continue to stick to a shell script.