At 02:08 PM 9/25/2002 -0400, psmith@xxxxxxx wrote:
The POSIX 1003.1-2001 "Shell and Utilities, Issue 6" spec (page 614,
lines 23699-23704) says:
If SHELL is defined in the makefile or is
specified on the command line, it shall replace the original value of
the SHELL macro, *but shall not affect the SHELL environment variable*.
I'm not a member of this list or a GNU make contributor, I'm just the one
who reported the bug. But FWIW and IMHO, the case for fixing this is pretty
clear. Start with the fact that there's no wiggle room in the standards;
POSIX and SUS are both perfectly clear. I'm aware that there are occasional
times when GNU will say "Look, POSIX is just stupid on such-and-such and
we're going to deliberately ignore it". But in this case POSIX isn't
stupid, or at least no one's come up with a reason why the nonstandard
behavior is better, and to date GNU make has been able to claim 100% POSIX
conformance.
Add to this the fact that any backward-compatibility problems can easily be
solved by prepending the 'export' directive to SHELL overrides, and as I
see it the case for a fix overwhelms any case against.
The compromise (wimpy) solution might be to fix it only in the presence of
.POSIX, but that's really ghetto-izing the fix since most users don't even
know about .POSIX let alone use it. Unless someone comes up with a reason
to turn a real fix into a pro forma fix, I wouldn't recommend going this way.
(David suggests that if you _explicitly_ export SHELL in your makefile,
then it would use that value in the environment).
Just to clarify, I don't make any suggestion regarding _use_ of SHELL from
the environment, as the current behavior is not broken. The sentence above
should read "... then it would _place_ that value in the environment". Of
course it's not much of a suggestion anyway, as this is just the standard
behavior of the export directive.
Thanks,
David Boyce
|
