Will Coleda wrote:
On Aug 23, 2007, at 10:31 PM, [EMAIL PROTECTED] wrote:
Someone actually reads my commit statements in a branch other than
trunk! Saints be praised!
First argument to split should be a regex, not a string. Made this
correction in three locations, which makes it more consistent with
gmake documentation.
This actually changes the behavior of this split (for good or ill):
From perldoc -f split:
> As a special case, specifying a PATTERN of space (' ') will
> split on white space just as "split" with no arguments does.
> A "split" on "/\s+/" is
> like a "split(' ')" except that any leading whitespace produces
> a null first field.
A feature of split so little used that it's the first I've heard of it
in 7+ years of Perl!
Nonetheless, I still think /\s+/ is the way to go. Here are the three
instances where I proposed changes to lib/Parrot/Configure/Step.pm's use
of split:
if (
$line =~ s{\$ \( notdir \s+ ([^)]+) \)}{
join (' ',
map { (File::Spec->splitpath($_))[2] }
split(/\s+/, $1)
)
# ...
if (
$line =~ s{\$ \( basename \s+ ([^)]+) \)}{
join (' ',
map {
my @split = File::Spec->splitpath($_);
$split[2] =~ s/\.[^.]*$//;
File::Spec->catpath(@split);
} split(/\s+/, $1)
# ...
if (
$line =~ s{\$ \( addprefix \s+ ([^,]+) \s* , \s*
([^)]+) \)}{
my ($prefix,$list) = ($1, $2);
join (' ',
map { $_ = $prefix . $_ }
split(/\s+/, $list)
)
In each, you have absorbed all possible leading whitespace in a \s+ or
\s*. So there's no downside to using /s+/. And there's an upside:
It's more forgiving than requiring that the developer type precisely one
wordspace characters between arguments.
Here's one of the test stanzas from the reconfigure branch's
t/configure/034-step.t that tests one of these gmake expansion statements:
{
my $tdir = tempdir( CLEANUP => 1 );
chdir $tdir or croak "Unable to change to temporary directory";
my $dummy = 'dummy';
open my $IN, '>', $dummy or croak "Unable to open temp file for
writing";
my $line = q{$(basename morgan/lefay/abra.ca.dabra src/foo.c
hacks)};
print $IN $line, "\n";
close $IN or croak "Unable to close temp file";
ok( genfile(
$dummy => 'CFLAGS',
expand_gmake_syntax => 1,
), "genfile() did transformation of 'make' 'basename' as expected");
unlink $dummy or croak "Unable to delete file after testing";
chdir $cwd or croak "Unable to change back to starting directory";
}
That's actually a hard-tab character (or two) between 'src/foo.c' and
'hacks'. This is consistent with the documentation I studied at
http://start.it.uts.edu.au/w/doc/solaris/gmake/make_8.html, which I
cited in an inline comment in that test file.
Further thoughts?
kid51
