[Reproducible-builds] [debhelper] 02/07: dh_strip: normalize ar file headers for reproducible builds

[Jérémy Bobbio]

This is an automated email from the git hooks/post-receive script.

lunar pushed a commit to branch pu/reproducible_builds
in repository debhelper.

commit 99adaf58da5f63065076d21818ce03ac3c7537a4
Author: Jérémy Bobbio <lunar at debian.org>
Date:   Sat Aug 30 02:24:57 2014 +0000

    dh_strip: normalize ar file headers for reproducible builds
    
    User id, group id, timestamps and file modes can get captured when creating
    static libraries. While this information is not useful while building software,
    it prevents build to be reproducible. Let's replace the data by what is
    written when `ar` is used in "deterministic" mode.
    
    Thanks to Niko Tyni for the Perl code.
---
 dh_strip | 61 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)

diff --git a/dh_strip b/dh_strip
index 516b6f2..c55f10f 100755
--- a/dh_strip
+++ b/dh_strip
@@ -8,6 +8,7 @@ dh_strip - strip executables, shared libraries, and some static libraries
 
 use strict;
 use File::Find;
+use Fcntl q/SEEK_SET/;
 use Debian::Debhelper::Dh_Lib;
 
 =head1 SYNOPSIS
@@ -28,6 +29,9 @@ strips each as much as is possible. (Which is not at all for debugging
 libraries.) In general it seems to make very good guesses, and will do the
 right thing in almost all cases.
 
+For static libraries, it will also normalize user id, group id, timestamp and
+file mode of the archive members to enable build reproducibility.
+
 Since it is very hard to automatically guess if a file is a
 module, and hard to determine how to strip a module, B<dh_strip> does not
 currently deal with stripping binary modules such as F<.o> files.
@@ -190,6 +194,61 @@ sub attach_debug {
 	doit($objcopy, "--add-gnu-debuglink", $debug_path, $file);
 }
 
+sub normalize_ar {
+	my $file=shift;
+
+	my $GLOBAL_HEADER= "!<arch>\n";
+	my $GLOBAL_HEADER_LENGTH=length $GLOBAL_HEADER;
+
+	my $FILE_HEADER_LENGTH=60;
+	my $FILE_MAGIC="`\n";
+
+	my $buf;
+
+	open(F, '+<', $file)
+		or die("failed to open $file for read+write: $!");
+
+	read F, $buf, $GLOBAL_HEADER_LENGTH;
+	die("Unable to find global header") if $buf ne $GLOBAL_HEADER;
+
+	while (1) {
+		my $file_header_start=tell F;
+		my $count=read F, $buf, $FILE_HEADER_LENGTH;
+		die "reading $file failed: $!" if !defined $count;
+		last if $count == 0;
+
+		# http://en.wikipedia.org/wiki/Ar_(Unix)
+		#from   to     Name                      Format
+		#0      15     File name                 ASCII
+		#16     27     File modification date    Decimal
+		#28     33     Owner ID                  Decimal
+		#34     39     Group ID                  Decimal
+		#40     47     File mode                 Octal
+		#48     57     File size in bytes        Decimal
+		#58     59     File magic                \140\012
+
+		die "Incorrect header length"
+			if length $buf != $FILE_HEADER_LENGTH;
+		die "Incorrect file magic"
+			if substr($buf, 58, length($FILE_MAGIC)) ne $FILE_MAGIC;
+
+		my $file_size = substr($buf, 48, 10);
+		seek F, $file_header_start + 16, SEEK_SET;
+
+		# mtime
+		syswrite F, sprintf("%-12d", 0);
+		# owner
+		syswrite F, sprintf("%-6d", 0);
+		# group
+		syswrite F, sprintf("%-6d", 0);
+		# file mode
+		syswrite F, sprintf("%-8o", 0644);
+
+		# move to next member
+		seek F, $file_header_start + $FILE_HEADER_LENGTH + $file_size, SEEK_SET;
+	}
+}
+
 foreach my $package (@{$dh{DOPACKAGES}}) {
 	my $tmp=tmpdir($package);
 
@@ -236,6 +295,7 @@ foreach my $package (@{$dh{DOPACKAGES}}) {
 
 	foreach (@static_libs) {
 		doit($strip,"--strip-debug",$_);
+		normalize_ar($_);
 	}
 }
 
@@ -248,5 +308,6 @@ This program is a part of debhelper.
 =head1 AUTHOR
 
 Joey Hess <joeyh at debian.org>
+Niko Tyni <ntyni at debian.org>
 
 =cut

-- 
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/reproducible/debhelper.git