Re: newline escape problem.

[David Boyce]

At 06:19 PM 10/29/2002 -0500, Paul D. Smith wrote:
> We already allow SHELL to contain whitespace, as in:
>
>   SHELL = /bin/sh -e
>
> (not portable but does work in some shells) so there is _SOME_ amount of
> parsing we have to do ...

What's the argument for allowing whitespace at all (aside from backward 
compatibility of course)? Most makes require SHELL to be a single word 
suitable for use as "arg0" in execl[p]() and in fact I've always assumed 
GNU make did the same thing. I don't think there will ever be a way to 
correctly parse SHELL into an array. What if it was

        SHELL := env FOO=BAR myscript -comment "this is what I propose to do"

If only there was a way for make to represent arrays; then we could 
contemplate an enhancement "MAKEARGV" (or whatever) which holds not a 
string requiring difficult parsing but a true array, e.g. in perl 
pseudo-syntax:

        MAKEARGV        :=      qw(my_program -flag1 -flag2 %)

where the command string gets plugged in in place of the '%' (adjust syntax 
to taste). This would solve a lot of the SHELL-customization issues in one 
fell swoop, but I'm not aware of a way to represent an array other than as 
a whitespace-separated string.

Next, a different but I hope sufficiently related topic: has the subject of 
implementing .ONESHELL for GNU make ever come up here? I see no RFE's for 
it in the bug database and google shows nothing (of course, in an apparent 
new "feature" google insists on telling me "ONESHELL" is a typo and 
separating it into "ONE SHELL"). Here's what SUS says:

     The default in some advanced versions of make is to group all the
     command lines for a target and execute them using a single shell
     invocation; the System V method is to pass each line individually
     to a separate shell. The single-shell method has the advantages in
     performance and the lack of a requirement for many continued lines.
     However, converting to this newer method has caused portability
     problems with many historical makefiles, so the behavior with the
     POSIX makefile is specified to be the same as that of System V. It
     is suggested that the special target .ONESHELL be used as an
     implementation extension to achieve the single-shell grouping for a
     target or group of targets.

I'd think it would be not so hard to implement, mostly a matter of joining 
lines with && between them, and a nice enhancement with a clear go-ahead 
from the relevant standards. So have I missed a deprecation of .ONESHELL or 
is it something that just hasn't come up?

-David