Re: parallel autotest [1/3]: Refactor testsuite driver loop.

[Eric Blake]

Ralf Wildenhues <Ralf.Wildenhues <at> gmx.de> writes:

> 
> rather than
>   2 3 11 failed, 4 5 10 passed unexpectedly
> 
> in the summary.  I didn't bother fixing those variables which we do not
> output; the question mark patterns are more, thus incur higher overhead.

Good explanation.  Perhaps a comment would be in order as a followup patch?

> 
> > Also, are '/*/' and
> > '/?/' portable patterns in sed, since those are regex metacharacters, or
> > do we need to play it safe and use '/\*/' and '/\?/'?
> 
> The '*' is special and needs escaping, fixed now, and pushed.  I have no
> idea how I could have overlooked that, but it seems that GNU and BSD sed
> do not interpret it as special when it's the first regex token.

And that is correct, according to POSIX: 
http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap09.html

"The asterisk shall be special except when used: ...
As the first character of an entire BRE (after an initial '^', if any)"


>  '?' is
> not special.

Oh right - I should have done a bit more research before asking :).  sed uses 
BRE, not ERE, so ? stands for itself (not to mention the autoconf manual 
already documents that \? in sed is not portable).  I guess my question was 
more along the lines of whether /*/ and /\*/ are universally interpreted as the 
same BRE like POSIX requires, or whether there are buggy sed out there 
where /*/ is treated as a syntax error where you HAVE to use /\*/ for 
portability.

-- 
Eric Blake