X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=dgit.git;a=blobdiff_plain;f=i18n-diff-auditor;h=6868ca7d2c8df16f3b6fee1169a3b7c04903e06c;hp=363103fccf01e7a4f7bfc47894330d6c56fe2066;hb=91ddd61e738d009851936647142dc94975d17882;hpb=6090adb68f0ab2b79adba51212bfe7b2dff0e6df

diff --git a/i18n-diff-auditor b/i18n-diff-auditor
index 363103fc..6868ca7d 100755
--- a/i18n-diff-auditor
+++ b/i18n-diff-auditor
@@ -4,6 +4,37 @@
 # Copyright (C)2018 Ian Jackson
 # GPLv3+, NO WARRANTY, see below.
 #
+#
+# Usage:
+#  something like this
+#  git-log -n1 -p | ./i18n-diff-auditor -D 2>&1 |less -j10 +/'^!.*'
+#
+# -D is for debug.  Currently only one level.
+#
+# Output is the relevant diff hunks, with each line prepended with
+# space for ok lines and ! for questionable ones, and with relevant
+# diff lines prepended with lines starting !! (and lines starting #
+# for debug output), so ovrall:
+#
+#   !! <message>   reasoning for subsequent questionable diff line(s)
+#   !+             diff line found to be questionable
+#   !-             diff line found to be questionable
+#    @@@ etc.      diff furniture
+#    +             diff line checked and ok
+#    -             diff line checked and ok
+#   # <stuff>      debug output (normally precedes relevant output)
+#
+# Changes are generally marked as ok if they correspond to a known
+# intended code change pattern.  (That includes changing error calls
+# to different error calls.)  If they don't correspond to any known
+# pattern, they are "questionable" and the first thing that doesn't
+# match the most common pattern is reported.
+#
+# Might be useful for projects other than dgit, provided it uses
+# the same gettext aliases (__ f_ i_) and similar error calls
+# (die, confess, fail).
+#
+#
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
 # the Free Software Foundation, either version 3 of the License, or
@@ -189,6 +220,7 @@ sub semiparse ($) {
 	    push @o, { T => 'specvar', E => $&, P => $& };
 	} elsif (!length) {
 	    last;
+	} elsif (s{^\#.*\n}{}) {
 	} else {
 	    m{^.{0,10}};
 	    die "cannot tokenise \`$&'";
@@ -220,8 +252,8 @@ our @analysed_x;
 our @analysed_y;
 
 sub analyse_chunk_core () {
+    $before //= '';
     die "plain deletion\n" unless defined $after;
-    die "plain insertion\n" unless defined $before;
     my @xs = semiparse $before;
     my @ys = semiparse $after;
     @analysed_x = @analysed_y = ();
@@ -236,6 +268,7 @@ sub analyse_chunk_core () {
     my $next_x = sub { $next_something->(\@xs, \@analysed_x, \$x, 'before'); };
     my $next_y = sub { $next_something->(\@ys, \@analysed_y, \$y, 'after' ); };
     our @y_expect_suffix = ();
+  ANALYSE:
     for (;;) {
 	while (my $e = shift @y_expect_suffix) {
 	    $next_y->();
@@ -249,6 +282,24 @@ sub analyse_chunk_core () {
 	next if $x->{E} eq 'sprintf' and $y->{E} eq 'f_';
 	next if $x->{E} eq 'die'     and $y->{E} eq 'confess';
 	next if $x->{E} eq 'die'     and $y->{E} eq 'fail';
+	foreach my $with_fh (qw(0 1)) {
+	    next unless $x->{E} eq 'printf';
+	    next unless $y->{E} eq 'print';
+	    next unless @xs >= $with_fh;
+	    next unless @ys >  $with_fh;
+	    if ($with_fh) {
+		next unless $xs[0]{E} eq $ys[0]{E};
+		next unless
+		    $xs[0]{E} =~ m{^[A-Z]+$} or
+		    $xs[0]{T} eq 'ident' && $xs[0]{E} =~ m{^\$};
+	    }
+	    next unless $ys[$with_fh]{E} eq 'f_';
+	    # yay!
+	    $next_x->() if $with_fh;
+	    $next_y->() if $with_fh;
+	    $next_y->(); # f_
+	    next ANALYSE;
+	}
 	if ($y->{E} eq '+'
 	    and @ys >= 3
 	    and $ys[0]{E} eq '('