# [PATCH] import-tars: properly import git-archive tarballs 6 messages from 2010-02-17 to 2010-02-19. Participants: Gabriel, Peter Krefting. Thread: https://gitlist.dev/t/22697 ## Gabriel, 2010-02-17 19:20 Subject: [PATCH] import-tars: properly import git-archive tarballs Message-ID: <1266434452-14532-1-git-send-email-g2p.code@gmail.com> URL: https://gitlist.dev/e/1266434452-14532-1-git-send-email-g2p.code%40gmail.com ``` git-archive adds a special entry to the archives it creates, pax_global_header, containing the SHA1 of the exported commit. import-tars.perl extracted it as a file, the top directory of the archive became a subdirectory, and files moved with every imported tarball. Now import-tars correctly ignores the comment. Signed-off-by: Gabriel --- contrib/fast-import/import-tars.perl | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/contrib/fast-import/import-tars.perl b/contrib/fast-import/import-tars.perl index 95438e1..a5170a1 100755 --- a/contrib/fast-import/import-tars.perl +++ b/contrib/fast-import/import-tars.perl @@ -108,6 +108,7 @@ foreach my $tar_file (@ARGV) } } print FI "\n"; + next if $typeflag eq 'g' && $name eq 'pax_global_header'; my $path; if ($prefix) { -- 1.7.0.rc2.31.g618d1 ``` ## Peter Krefting, 2010-02-18 09:40 Subject: Re: [PATCH] import-tars: properly import git-archive tarballs Message-ID: URL: https://gitlist.dev/e/alpine.DEB.2.00.1002181037340.11095%40ds9.cixit.se In-Reply-To: <1266434452-14532-1-git-send-email-g2p.code@gmail.com> ``` Gabriel: > git-archive adds a special entry to the archives it creates, > pax_global_header, containing the SHA1 of the exported commit. Interesting use-case. > Now import-tars correctly ignores the comment. In the case where you actually would care, perhaps it should be added as a comment to the default commit message? Something like "Imported from $tar_file, created from commit $commit.". -- \\// Peter - http://www.softwolves.pp.se/ ``` ## Gabriel, 2010-02-18 13:00 Subject: Re: [PATCH] import-tars: properly import git-archive tarballs Message-ID: <20100218140002.14f79853@gmail.com> URL: https://gitlist.dev/e/20100218140002.14f79853%40gmail.com In-Reply-To: <1266434452-14532-1-git-send-email-g2p.code@gmail.com> ``` > Gabriel: > > git-archive adds a special entry to the archives it creates, > > pax_global_header, containing the SHA1 of the exported commit. > > Interesting use-case. My use case really was a simple “import a range of tarballs for bisecting”. That they were from git was a surprise, but I don't have access to that repo, which for all I know is full of junk and not meant for publication, so I'm not trying to get the original history. I just want importing to work on tarballs found in the wild. > > Now import-tars correctly ignores the comment. > > In the case where you actually would care, perhaps it should be added > as a comment to the default commit message? Something like "Imported > from $tar_file, created from commit $commit.". Someone with stronger perl-fu to parse the hash comment could print on stderr (keep me CC-ed if you do that): This tarball was generated by git, from commit id $SHA1. It will serve as a hint when the person doing the import has access to the original history. I don't think there's a point in keeping a sha1 when there's no access; keeping it in the history would confuse people, or require sticking a longer explanation to each of these commits. ``` ## Peter Krefting, 2010-02-18 13:24 Subject: [PATCH] import-tars: Make it possible to include the original commit ID Message-ID: <20100218132709.8CEBA2FC07@perkele> URL: https://gitlist.dev/e/20100218132709.8CEBA2FC07%40perkele In-Reply-To: <20100218140002.14f79853@gmail.com> ``` Add an option --readpax which makes import-tars read the commit ID stored by git-archive, adding it to the default commit message. Signed-off-by: Peter Krefting --- > Someone with stronger perl-fu to parse the hash comment could print on > stderr (keep me CC-ed if you do that): > This tarball was generated by git, from commit id $SHA1. Actually, having this information could be useful, for instance to create a "history-skipping" repository, and use bisect in that. When one finds which released version causes the problem, one can bisect in the original repository, using the commit ID indicated. This applies on top of your patch. contrib/fast-import/import-tars.perl | 25 +++++++++++++++++++++++-- 1 files changed, 23 insertions(+), 2 deletions(-) diff --git a/contrib/fast-import/import-tars.perl b/contrib/fast-import/import-tars.perl index a5170a1..3451309 100755 --- a/contrib/fast-import/import-tars.perl +++ b/contrib/fast-import/import-tars.perl @@ -14,14 +14,20 @@ ## ## echo 'This is the commit message' > myfile.tar.bz2.msg ## perl import-tars.perl --metainfo=msg myfile.tar.bz2 +## +## Use --readpax to read the pax_global_header generated by git archive. +## The commit ID stored in the header will be appended to the default +## generated commit message for the imported tarball. If the parameter +## is not given, the pax_global_header is ignored. use strict; use Getopt::Long; my $metaext = ''; +my $readpax; -die "usage: import-tars [--metainfo=extension] *.tar.{gz,bz2,lzma,xz,Z}\n" - unless GetOptions('metainfo=s' => \$metaext) && @ARGV; +die "usage: import-tars [--metainfo=extension] [--readpax] *.tar.{gz,bz2,lzma,xz,Z}\n" + unless GetOptions('metainfo=s' => \$metaext, 'readpax' => \$readpax) && @ARGV; my $branch_name = 'import-tars'; my $branch_ref = "refs/heads/$branch_name"; @@ -61,6 +67,7 @@ foreach my $tar_file (@ARGV) my $author_time = 0; my $next_mark = 1; my $have_top_dir = 1; + my $orig_commitid = ''; my ($top_dir, %files); while (read(I, $_, 512) == 512) { @@ -95,6 +102,16 @@ foreach my $tar_file (@ARGV) $size = oct $size; $mtime = oct $mtime; next if $typeflag == 5; # directory + if (defined $readpax && $typeflag eq 'g' && $name eq 'pax_global_header') + { + while ($size > 0 && read(I, $_, 512) == 512) { + if (/52 comment=([0-9a-f]{40})/) + { + $orig_commitid = $1; + } + } + next; + } print FI "blob\n", "mark :$next_mark\n"; if ($typeflag == 2) { # symbolic link @@ -154,6 +171,10 @@ foreach my $tar_file (@ARGV) close MSG; } } + elsif ($orig_commitid ne '') + { + $commit_msg .= "\n\nThis tarball was generated by git, from commit id\n$orig_commitid."; + } print FI < URL: https://gitlist.dev/e/1266505333-20286-1-git-send-email-g2p.code%40gmail.com In-Reply-To: <20100218132709.8CEBA2FC07@perkele> ``` From: Peter Krefting Add an option --readpax which makes import-tars read the commit ID stored by git-archive, adding it to the default commit message. Signed-off-by: Peter Krefting Signed-off-by: Gabriel --- That's a sensible patch, thanks. A small correction; when both --readpax and --metainfo=msg are passed, incorporate both message and original sha1. I amended your commit. contrib/fast-import/import-tars.perl | 27 ++++++++++++++++++++++++--- 1 files changed, 24 insertions(+), 3 deletions(-) diff --git a/contrib/fast-import/import-tars.perl b/contrib/fast-import/import-tars.perl index a5170a1..f41480e 100755 --- a/contrib/fast-import/import-tars.perl +++ b/contrib/fast-import/import-tars.perl @@ -14,14 +14,20 @@ ## ## echo 'This is the commit message' > myfile.tar.bz2.msg ## perl import-tars.perl --metainfo=msg myfile.tar.bz2 +## +## Use --readpax to read the pax_global_header generated by git archive. +## The commit ID stored in the header will be appended to the default +## generated commit message for the imported tarball. If the parameter +## is not given, the pax_global_header is ignored. use strict; use Getopt::Long; my $metaext = ''; +my $readpax; -die "usage: import-tars [--metainfo=extension] *.tar.{gz,bz2,lzma,xz,Z}\n" - unless GetOptions('metainfo=s' => \$metaext) && @ARGV; +die "usage: import-tars [--metainfo=extension] [--readpax] *.tar.{gz,bz2,lzma,xz,Z}\n" + unless GetOptions('metainfo=s' => \$metaext, 'readpax' => \$readpax) && @ARGV; my $branch_name = 'import-tars'; my $branch_ref = "refs/heads/$branch_name"; @@ -61,6 +67,7 @@ foreach my $tar_file (@ARGV) my $author_time = 0; my $next_mark = 1; my $have_top_dir = 1; + my $orig_commitid = ''; my ($top_dir, %files); while (read(I, $_, 512) == 512) { @@ -95,6 +102,16 @@ foreach my $tar_file (@ARGV) $size = oct $size; $mtime = oct $mtime; next if $typeflag == 5; # directory + if (defined $readpax && $typeflag eq 'g' && $name eq 'pax_global_header') + { + while ($size > 0 && read(I, $_, 512) == 512) { + if (/52 comment=([0-9a-f]{40})/) + { + $orig_commitid = $1; + } + } + next; + } print FI "blob\n", "mark :$next_mark\n"; if ($typeflag == 2) { # symbolic link @@ -124,7 +141,7 @@ foreach my $tar_file (@ARGV) $have_top_dir = 0 if $top_dir ne $1; } - my $commit_msg = "Imported from $tar_file."; + my $commit_msg = "Imported from $tar_file.\n"; my $this_committer_name = $committer_name; my $this_committer_email = $committer_email; my $this_author_name = $author_name; @@ -154,6 +171,10 @@ foreach my $tar_file (@ARGV) close MSG; } } + if ($orig_commitid ne '') + { + $commit_msg .= "\nThis tarball was generated by git, from commit id\n$orig_commitid."; + } print FI < URL: https://gitlist.dev/e/alpine.DEB.2.00.1002190911140.5644%40ds9.cixit.se In-Reply-To: <1266505333-20286-1-git-send-email-g2p.code@gmail.com> ``` Gabriel: > A small correction; when both --readpax and --metainfo=msg are passed, > incorporate both message and original sha1. I amended your commit. My initial thought was to let have the --metainfo message override the auto-generated message in its entirety, but I have no strong feelings either way, so i am OK with this change. -- \\// Peter - http://www.softwolves.pp.se/ ```