[Libguestfs] [PATCH 2/8] v2v: add basic support for the "deb" package manager

[Tomáš Golembiovský]

On Mon,  8 Aug 2016 18:38:49 +0200
Pino Toscano <ptoscano at redhat.com> wrote:

> Implement the 'remove', 'file_list_of_package', and 'file_owner' methods
> of the Linux module for the "deb" package manager (dpkg basically, on
> Debian and derived distributions).
> 
> Also allow it for the main conversion code.
> ---
>  v2v/convert_linux.ml |  2 +-
>  v2v/linux.ml         | 23 +++++++++++++++++++++++
>  2 files changed, 24 insertions(+), 1 deletion(-)
> 
> diff --git a/v2v/convert_linux.ml b/v2v/convert_linux.ml
> index 4b1ce99..65796d6 100644
> --- a/v2v/convert_linux.ml
> +++ b/v2v/convert_linux.ml
> @@ -79,7 +79,7 @@ let rec convert ~keep_serial_console (g : G.guestfs) inspect source rcaps =
>      | "sles" | "suse-based" | "opensuse" -> `SUSE_family
>      | _ -> assert false in
>  
> -  assert (inspect.i_package_format = "rpm");
> +  assert (inspect.i_package_format = "rpm" || inspect.i_package_format = "deb");
>  
>    (* We use Augeas for inspection and conversion, so initialize it early. *)
>    Linux.augeas_init g;
> diff --git a/v2v/linux.ml b/v2v/linux.ml
> index ed639c1..5713f64 100644
> --- a/v2v/linux.ml
> +++ b/v2v/linux.ml
> @@ -48,6 +48,10 @@ and do_remove g inspect packages =
>    assert (List.length packages > 0);
>    let package_format = inspect.i_package_format in
>    match package_format with
> +  | "deb" ->
> +    let cmd = [ "dpkg"; "--purge" ] @ packages in
> +    let cmd = Array.of_list cmd in
> +    ignore (g#command cmd);
>    | "rpm" ->
>      let cmd = [ "rpm"; "-e" ] @ packages in
>      let cmd = Array.of_list cmd in
> @@ -61,6 +65,12 @@ let file_list_of_package (g : Guestfs.guestfs) inspect app =
>    let package_format = inspect.i_package_format in
>  
>    match package_format with
> +  | "deb" ->
> +    let cmd = [| "dpkg"; "-L"; app.G.app2_name |] in
> +    debug "%s" (String.concat " " (Array.to_list cmd));
> +    let files = g#command_lines cmd in
> +    let files = Array.to_list files in
> +    List.sort compare files
>    | "rpm" ->
>      (* Since RPM allows multiple packages installed with the same
>       * name, always check the full ENVR here (RHBZ#1161250).
> @@ -98,6 +108,19 @@ let file_list_of_package (g : Guestfs.guestfs) inspect app =
>  let rec file_owner (g : G.guestfs) inspect path =
>    let package_format = inspect.i_package_format in
>    match package_format with
> +  | "deb" ->
> +      let cmd = [| "dpkg"; "-S"; path |] in
> +      debug "%s" (String.concat " " (Array.to_list cmd));
> +      let lines =
> +        try g#command_lines cmd

This is not good. dpkg-query command behaves differently from rpm.
First, the returned packages are all on one line separated by commas.
What I came up with is this:

      let cmd = [| "dpkg-query"; "-S"; path |] in
      debug "%s" (String.concat " " (Array.to_list cmd));
      (try
         let pkg = g#command cmd in
         (* What we get is a string of form "pkg1, pkg2:arch, pkg3: /path" *)
         let len = String.length pkg in
         let rec loop i =
             if i >= len then
               (* This is fishy. Internal bug? *)
               error (f_"internal error: file_owner: failed to process string '%s'") pkg
             else if pkg.[i] = ' ' &&
               (pkg.[i-1] = ':' || pkg.[i-1] = ',') then
                 String.sub pkg 0 (i-1)
             else loop (i+1)
         in loop 1
       with Guestfs.Error msg as exn ->
         if String.find msg "no path found matching pattern" >= 0 then
           raise Not_found
         else
           raise exn
      )

The other concern is, whether we should escape the argument somehow. The
fact is that dpkg-query accepts a glob expression, not just a simple
path. If you provide it with a pattern you will get all the matching
paths, each on a single line (prefixed by the package list).

This might not be an issue. Depends on how we use the method. Thoughts?


> +        with Guestfs.Error msg as exn ->
> +          if String.find msg "no path found matching pattern" >= 0 then
> +            raise Not_found
> +          else
> +            raise exn in
> +      if Array.length lines = 0 then
> +        error (f_"internal error: file_owner: dpkg command returned no output");
> +      lines.(0)
>    | "rpm" ->
>        (* Although it is possible in RPM for multiple packages to own
>         * a file, this deliberately only returns one package.


-- 
Tomáš Golembiovský <tgolembi at redhat.com>