Re: ./init.sh: line 139: `test-acos': not a valid identifier

[Jim Meyering]

Eric Blake wrote:
> According to Jim Meyering on 2/18/2010 1:34 AM:
>> Simon Josefsson wrote:
>>> ./init.sh: line 139: `test-acos': not a valid identifier
>>> FAIL: test-pread.sh
>> That failure is because we cannot create a function name containing a "-"
>> using that shell.  This is another portability hurdle that we should
>> be able to overcome with what Eric proposed: have init.sh reinvoke
>> a better shell.
>
> Another alternative would be changing all test scripts to use _ rather
> than -, so as to avoid issues with requiring shell extensions.  But I'm

Remember that the goal of this exercise (for me at least)
was to avoid letting the requirement for .exe support impose
inordinate ugliness and maintenance overhead in all of our
hundreds of test scripts.  If we let this infrastructure dictate
the form of our file names (changing from hyphenated to inferior-imho
underscore-using names), we've admitted defeat ;-)

> not fond of that idea.  On the other hand, it seems awkward that we are
> basically requiring bash (and not POSIX) to run on any platform where
> $EXEEXT is important (cygwin, mingw/wine, ...).

If one of those systems must refrain from running some script-based
tests because it lacks a capable shell, that's not a big deal.
If someone cares, they will install bash before the next run.