Re: Problems with echo. (echo period)

[Eli Zaretskii]

> Date: Mon, 23 Apr 2007 07:20:53 -0700 (PDT)
> From: Aaron Shatters <address@hidden>
> Cc: address@hidden
> 
> >I remain unconvinced that we should ``fix'' this just because this
> >fundamental incompatibility between shells sometimes causes trouble in
> >Makefiles that assume Posix `echo' where it isn't guaranteed to exist.
> 
> The problem is that Make is *not* invoking the shell built-in

I was talking about the situation where `echo' is called with an empty
command line.  In that case, the shell does get called.

As for the "echo." case, it's just a bizarre consequence of cmd
command-line parsing (there are many more manifestations of that, and
others have pointed out some of them).  I don't think there's a very
large demand for supporting that in Make, and even if there is, I
don't plan on working on that any time soon.

> I am confused as to what level of support Make is trying to achieve.  There 
> are !unixy_shell conditions and dos cmd references all over the code.  What 
> are the purposes of these if not to handle "fundamental incompatibilities 
> between shells".  I don't understand where the point of disagreement is.  I 
> understand that this is not a bug in Make, but, in my opinion, this is a 
> perfectly reasonable enhancement.  Make knows about the existence of the 
> windows shell 'cmd.exe'.  It knows what the shell built-ins are, and it 
> already handles this shell specially (for example, sh_cmds_dos in job.c).  
> What reasoning is there for excluding this functionality, when all of the 
> other specifics associated with this shell are handled?

My reasoning is that supporting such weirdo parsing is a maintenance
burden.  The way Make parses the commands now cannot be easily
modified to cater to these strange corners in cmd.

You can always say `cmd /c "echo."', or use some special character,
such as redirection or pipe, if you need to force Make to call the
shell.