Re: *_DATA containing blanks

[Ralf Corsepius]

On Thu, 2003-07-24 at 09:59, Alexandre Duret-Lutz wrote:
> >>> "Ralf" == Ralf Corsepius <address@hidden> writes:
> 
> [...]
> 
>  Ralf> if test -n '$(%DIR%_%PRIMARY%)'; then \
>  Ralf> for p in ""$(%DIR%_%PRIMARY%); do
> 
>  Ralf> Seems to work with Solaris 2.7 (/bin/sh + /usr/ccs/bin/make)
>  Ralf> and RH-9 (bash + gmake).
> 
> You'll kill me, 
I have to admit, now considering this thought ;)

> but I've just remembered a reason why Automake uses
>   list='$(VAR)'; for i in $$list; 
> over 
>   for i in $(VAR); 
> even when VAR is known not to be empty.
> 
> It's because we must support filenames containing `$'.  Java
> users need this,

Who needs java ? ;-)

>  and AFAIU they don't have to possibility to
> rename their files.  There is a test case for this: dollar.test.
I can easily contribute a test case to demonstrate my problem.

=> we have contradicting demands, fixing one issue breaks the other.

=> Automake's has a fundamental problem

I am inclined to read this as an indication that we are playing with
symptoms, while the actual origin of the problems is elsewhere.

Just a thought: Don't automake parse the contents of make-variables? 
Why doesn't it expand these variables and use expanded loops instead of
trying to use these vars directly? This would also give automake the
opportunity to quote sick vars (like those using $ or ' ') correctly?


For the moment, I'll give in and have to conclude that automake is not
applicable to my problem.

Ralf