[clue-tech] Portability of shell expressions in a Makefile

[Keith Hellman]

On Sun, May 08, 2005 at 10:32:44PM -0600, Matt Gushee wrote:
> Hello, all--
> 
> I am using the following to set a variable in a Makefile:
> 
> 	SOLID=$(shell \
> 		version=`ocamlc -version`; \
> 		major=$${version:0:4}; \
> 		minor=$${version:5}; \
> 		if [ $$major = "3.08" ] && [ $$minor ] && \
> 			[ $$minor -ge 3 ]; then \
> 			echo "-DSOLID_RELIEF"; \
> 		else \
> 			echo ""; \
> 		fi )
> 
> This has the desired effect in bash. Can anyone tell me if it will work 
> with other Bourne-compatible shells? And while we're at it, how about 
> non-GNU make implementations?

I can't comment on non-GNU make implementations cause I've haven't used
one for years.  I wouldn't be surprised that most people do have it
installed if they use any amount of FOSS software.

According to
  http://www.linuxselfhelp.com/gnu/bash/html_chapter/bashref_4.html
the ${var:offset:len} is not a standard in sh.  You could avoid
worrying about it (and require people to at least have bash
on their systems) by doing this:

SOLID=$(shell bash -c '\
	version=3.08.6 ; \
	major=$${version:0:4}; \
	minor=$${version:5}; \
	if [ $$major = "3.08" ] && [ $$minor ] && \
		[ $$minor -ge 3 ]; then \
		echo >&2 yeah $$major $$minor ; \
		echo "-DSOLID_RELIEF"; \
	else \
		echo >&2 boohoo $$major ; \
		echo >&2 boohoo $$minor ; \
		echo ""; \
	fi' )

This above snippit worked on my machine (debian sarge, gmake 3.80).
Note that worked around not having ocaml installed :^)

HTH 

-- 
Keith Hellman                             #include <disclaimer.h>
khellman at mcprogramming.com                from disclaimer import standard
public key @ www.mcprogramming.com


"While it might seem that a simple increment operator is an atomic
operation, there's no gaurantee that it is.  It's actually possible for
Thread 1 to update half of a 32-bit x while Thread 2 reads the full 32
bits, getting a mishmash instead of a valid integer.  (And that's only
one thing that can go wrong.  Another is that compiler optimization might
leave the integer in a register.  You really can't ever let threads
simultaneously access data without protection.)"

-- Marc J. Rochkind, *Advanced Unix Programming*, Second Ed.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://cluedenver.org/pipermail/clue-tech/attachments/20050508/a2ffb6c9/attachment.bin