Hi James,
On Tuesday 25 May 2010 17:55:53 James Vega wrote:
> I noticed via one of your other patches what looks to be a bad check for file
> redirection.
>
> qr'(<\&|>\&)\s*((-|\d+)[^\s;|)}`&\\\\]|[^-\d\s]+)' =>
> q<should be \>word 2\>&1>,
>
> This will have false positives for things like the following:
>
> $as_echo "$as_me:${as_lineno-$LINENO}: error: $1" >&$3
>
> That is, we should allow for the results of an expansion to be used for
> “word” (according to SUS's[0] use of the term).
The problem is that if $x doesn't contain a fd number then we would miss those
cases. I think we could allow \$\d to hide the autoconf false positives.
> These changes introduces a false positive wrt here strings.
>
> possible bashism in src/auto/configure line 4650 (<<< here string):
> { $as_echo "$as_me:${as_lineno-$LINENO}: result: >>> Perl >
> 5.6 with 5.5 threads cannot be used <<<" >&5
This one comes from the set of "it could already be triggered but the code was
simply ignored." Running an old checkbashisms against that line produces the
FP too.
But thanks, I've just added it to the testsuite.
The check for function names is now more likely to be more (FP-wise) visible
on sed scripts too.
Cheers,
--
Raphael Geissert - Debian Developer
www.debian.org - get.debian.net
--
To unsubscribe, send mail to [email protected].
