On 01/14/2010 06:42 PM, Junio C Hamano wrote:
quoted text
> Eric Blake<ebb9@byu.net>  writes:
>
>> Michael Haggerty<mhagger<at>  alum.mit.edu>  writes:
>>
>>>   	current_sha1=$(git rev-parse --verify HEAD)
>>> -	if test "$no_ff$current_sha1" = "$parent_sha1"; then
>>> +	if test -z "$no_ff" -a "$current_sha1" = "$parent_sha1"
>>
>> 'test cond1 -a cond2' is not portable.  Use 'test cond1&&  test cond2'.
>
> I avoid "test -a/-o" myself without even thinking (I am from old-school),
> but at the same time I thought the progress in the world made such caution
> obsolescent.
>
> Not so.  Even though POSIX.1 lists -a/-o as options to "test", they are
> marked "Obsolescent XSI" ("Strictly Conforming POSIX Applications and
> Strictly Conforming XSI Applications shall not use obsolescent features").

The reason for this is that the precedence rules were never well 
specified, and this made many sane-looking uses of "test -a/-o" problematic.

For example, if $x is "=", these work according to POSIX (it's not 
portable, but in practice it's okay):

   $ test -z "$x"
   $ test -z "$x" && test a = b

but this doesn't

   $ test -z "$x" -a a = b
   bash: test: too many arguments

because it groups "test -n = -a" and is left with "a = b".

Similarly, if $x is "-f", these

   $ test "$x"
   $ test "$x" || test c = d

correctly adds an implicit "-n", but this fails:

   $ test "$x" -o c = d
   bash: test: too many arguments

If anybody cleans up git's usage of test -a/-o, feel free to cut'n'paste 
the above into the commit messages.

Paolo
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html