[PATCH] Use printf(1) instead of using bash-specific shell code

[Giorgos Keramidas]

On 2007-03-17 15:18, Matt Mackall <mpm at selenic.com> wrote:
> On Sat, Mar 17, 2007 at 10:23:48PM +0200, Giorgos Keramidas wrote:
> > # HG changeset patch
> > # User Giorgos Keramidas <keramida at ceid.upatras.gr>
> > # Date 1167507225 -7200
> > # Node ID 9b36d9e4f429b741b445fb6a97bbad37fcd6c92a
> > # Parent  641c3bb47e839842f05e28a86473efc6555a49ac
> > Use printf(1) instead of using bash-specific shell code.
> 
> Perhaps we should have a test that greps the other test cases for
> non-portable constructs.

That's a very interesting idea :)

grep may be inadequate, but I'm trying to periodically run the tests in
various operating systems, including:

  * FreeBSD 6-STABLE
  * FreeBSD 7.0-CURRENT
  * Fedora Core Linux
  * Debian GNU/Linux
  * Solaris 9
  * Solaris 10 (currently Update 3)

Some times, this catches constructs which are non-portable to *all* the
systems in question.  But a non-portable construct is not always a bug,
so automating this sort of checking may be kind of tricky.