-pristine-tar (0.23) UNRELEASED; urgency=low
-
- * pristine-gz: use xdelta to check if a file was closely, but not completly,
- reproduced. This is not yet used except as a debugging aid when
- no variants succeed.
+pristine-tar (1.00) UNRELEASED; urgency=low
+
+ * pristine-gz: Fall back to storing a binary delta, in the rare
+ cases where the file cannot be 100% reproduced. A warning message
+ is printed if the binary delta is not relatively small.
+ * Allows generating quite small deltas for php-geoip,
+ xsupplicant, libgraphics-colornames-perl, and
+ Ricoh-Aficio_2020D-Postscript.ppd.gz -- zgz gets very close to the
+ originals for all of these.
+ Closes: #518972, #506874
+ * For other things the deltas are not small, but only a few hundred
+ known tarballs fail that badly, and the total space wasted by all
+ of the oversized deltas for them would only be 100 mb. This seems
+ an accepable tradeoff to be able to reliably use it on anything.
+ Closes: #475778, #509703, #509707, #515256, #515331
+ * Increase gz delta version number to 3.0 if a binary delta file
+ is included.
-- Joey Hess <joeyh@debian.org> Mon, 13 Apr 2009 15:12:44 -0400
In a few cases post-build fixups are also done to ensure that the gz is
identical to the original.
-This approach will work in the vast majority of cases. If it doesn't work,
-no delta will be generated. One example of a case it cannot currently
-support is a gz file that has been produced by appending together multiple
-gz files.
+This approach will work for about 99.5% of cases. One example of a case it
+cannot currently support is a gz file that has been produced by appending
+together multiple gz files.
+
+For the few where it doesn't work, a binary diff will be included in the
+delta between the closest regneratable gz file and the original. In
+the worst case, the diff will include the entire content of the original
+gz file, resulting in a larger than usual delta. If the delta is much
+larger than usual, pristine-gz will print a warning.
If the delta filename is "-", pristine-gz reads or writes it to stdio.
}
my $origsize=(stat($orig))[7];
- my $bestvariant;
- my $bestsize=$origsize;
+ my ($bestvariant, $bestsize);
foreach my $variant (@try) {
doit_redir($tempin, $tempout, 'zgz', @$variant, @extraargs, '-c');
if (!comparefiles($orig, $tempout)) {
# success
- return $name, $timestamp, @$variant;
+ return $name, $timestamp, undef, @$variant;
}
else {
# generate a binary delta and see if this is the
# best variant so far
- my $ret=system("xdelta delta -0 --pristine $tempout $orig $tempdir/delta 2>/dev/null") >> 8;
+ my $ret=system("xdelta delta -0 --pristine $tempout $orig $tempdir/tmpdelta 2>/dev/null") >> 8;
# xdelta exits 1 on success
if ($ret == 1) {
- my $size=(stat("$tempdir/delta"))[7];
- if ($size < $bestsize) {
+ my $size=(stat("$tempdir/tmpdelta"))[7];
+ if (! defined $bestsize || $size < $bestsize) {
$bestvariant = $variant;
$bestsize=$size;
+ rename("$tempdir/tmpdelta", "$tempdir/bestdelta") || die "rename: $!";
}
}
}
}
- print STDERR "pristine-gz failed to reproduce build of $orig\n";
- if (defined $bestvariant) {
- my $percentsaved=(($origsize-$bestsize)/$origsize*100);
- print STDERR "(best variant, saving $percentsaved%: @$bestvariant)\n";
+ # Nothing worked perfectly, so use the delta that was generated for
+ # the best variant
+ my $percentover=100 - int (($origsize-$bestsize)/$origsize*100);
+ debug("Using delta to best variant, bloating $percentover%: @$bestvariant");
+ if ($percentover > 10) {
+ print STDERR "warning: pristine-gz cannot reproduce build of $orig; ";
+ if ($percentover >= 100) {
+ print STDERR "storing entire file in delta!\n";
+ }
+ else {
+ print STDERR "storing $percentover% size diff in delta\n";
+ }
+ print STDERR "(Please consider filing a bug report so the delta size can be improved.)\n";
}
- print STDERR "Please file a bug report.\n";
- exit 1;
+ return $name, $timestamp, "$tempdir/bestdelta", @$bestvariant;
}
sub gengz {
open (IN, "$tempdir/version") || die "delta lacks version number ($!)";
my $version=<IN>;
- if ($version >= 3) {
+ if ($version >= 4) {
die "delta is version $version, not supported\n";
}
close IN;
chomp $timestamp;
close IN;
- doit_redir("$file", "$file.gz", "zgz", @params, "-T", $timestamp, "-F", "$filename", "-c");
+ my @zgz=("zgz", @params, "-T", $timestamp, "-F", "$filename", "-c");
+
+ if (-e "$tempdir/delta") {
+ doit_redir($file, "$tempdir/$file.gz", @zgz);
+ doit("xdelta", "patch", "--pristine", "$tempdir/delta", "$tempdir/$file.gz", "$file.gz");
+ }
+ else {
+ doit_redir("$file", "$file.gz", @zgz);
+ }
}
sub gendelta {
my @files=qw(version type params filename timestamp);
- my ($filename, $timestamp, @params)=
+ my ($filename, $timestamp, $xdelta, @params)=
reproducegz($gzfile, $tempdir, "$tempdir/test");
-
+
open(OUT, ">", "$tempdir/version") || die "$!";
- print OUT "2.0\n";
+ print OUT (defined $xdelta ? "3.0" : "2.0")."\n";
close OUT;
open(OUT, ">", "$tempdir/type") || die "$!";
print OUT "gz\n";
open(OUT, ">", "$tempdir/timestamp") || die "$!";
print OUT "$timestamp\n";
close OUT;
+ if (defined $xdelta) {
+ rename($xdelta, "$tempdir/delta") || die "rename: $!";
+ push @files, "delta";
+ }
doit("tar", "czf", $delta, "-C", $tempdir, @files);
projects / tools / pristine-tar.git / commitdiff