Re: [Bug-wget] bootstrap problem

[Jim Meyering]

On Tue, May 5, 2015 at 2:50 PM, Ángel González <address@hidden> wrote:
> On 05/05/15 12:45, Dagobert Michelsen wrote:
>>
>> Hi Darshit,
>>
>> Am 05.05.2015 um 12:23 schrieb Darshit Shah<address@hidden>:
>>>
>>> So, is the issue because some platforms are shipping a non-compatible
>>> version of sh, or is the script incorrectly using bash while only
>>> claiming to use sh?
>>
>> The problem is the evaluation with $( ):
>>    bootstrap: syntax error at line 88: `me_=$’ unexpected
>>
>> The Solaris Bourne Shell is really one of the „very oldest“
>> non-POSIX-compatible
>> Bourne shells:
>>    http://mywiki.wooledge.org/BashFAQ/082
>>
>> Last time I talked with Jim Meyering about the problem (unfortunately
>> off-list)
>> he said that if you bootstrap stuff you are a developer which is expected
>> to have
>> bash at hand. Here, bash is considered similar to autotools which you also
>> need
>> only to bootstrap.
>>
>> Best regards
>
>
> I don't think that we should spend any effort attempting to support
> non-POSIX shells
> (Solaris sh seems the only midly common under that category).
> That said, it would be nice to have a nicer error message.

Happy to reiterate :-)
Please don't waste time on shells that don't support $(...).
I agree that it is important to die early when the selected shell is inadequate.
I prefer to use the approach taken in gnulib's init.sh, which is used in most
of coreutils' tests:
http://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/init.sh