Bug#690014: getting rid of the config.guess/sub problem when bootstrapping new ports/systems
Wookey <wookey@wookware.org> writes:
> I had a go at this yesterday, updating some files but then realising
> that I didn't understand how lintian is put together. The current tests
> distinguish between 'out of date' (currently 2006) and 'ancient'
> (currently 2001), but don't apply equally to .guess and .sub, which
> seems odd to me. I thought that june 2012 (ie including aarch64) was a
> good new definition of 'out of date', but I'm not sure what 'ancient'
> should be moved up to. 2005 ish?
ancient means that the package will probably have problems with release
architectures. outdated means that the package will probably work on all
release architectures but will have problems with ports. (We should say
that more explicitly somewhere.)
They should apply equally to config.guess and config.sub so far as I can
tell. The code is in checks/cruft and is:
} elsif ($name =~ m,^(.+/)?config.(?:guess|sub)$, and not $atdinbd) {
my $b = basename $name;
open F, '<', $b or fail "can't open $name: $!";
while (<F>) {
last if $. > 10; # it's on the 6th line, but be a bit more lenient
if (/^(?:timestamp|version)='((\d+)-(\d+).*)'$/) {
my ($date, $year, $month) = ($1, $2, $3);
if ($year < 2004) {
tag 'ancient-autotools-helper-file', $name, $date;
} elsif (($year < 2006) or ($year == 2006 and $month < 6)) {
tag 'outdated-autotools-helper-file', $name, $date;
}
}
}
close F;
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: