Failed to use Tar.read with Tar_gz.in_gzipped

[BChabanne]

Combining Tar.read with Tar_gz.in_gzipped is failing.

I think this is because the "gzip buffer" is not being used.
Maybe there is a missing function read_through_gz here ?

ocaml-tar/lib/tar_gz.ml

Line 116 in 4215ff0

| Tar.Read _len as v -> v

Or maybe it is not expected to use Tar.read and only Tar.really_read shoud be used ? 

I tried to get a minimal failing script. The following program gives me this output.
read_rec is calling reccursively Tar.read until n bytes have been read. I guess it should have the same behaviour with Tar.really_read

read_really tar...               
file:test len:8
ok
read_rec tar...
file:test len:8
ok
read_really targz...
file:test len:8
ok
read_rec targz...
error: Unexpected end of file

(* bash :
    echo content > test
    tar -cf test.tar test
    tar -czf test.tar.gz test
*)
let decompress () =
  let ( let* ) = Tar.( let* ) in

  let print_err = function
    | Ok () -> Format.printf "ok\n"
    | Error `Eof -> Format.printf "error: end of file\n"
    | Error (`Gz msg) -> Format.printf "error: gz %s\n" msg
    | Error (#Tar_unix.error as err) ->
        Format.printf "error: %a\n" Tar_unix.pp_error err
  in

  let fold f ?global:_ (header : Tar.Header.t) () =
    let* content = f (Int64.to_int header.file_size) in
    Format.printf "file:%s len:%d\n" header.file_name (String.length content);
    Tar.return (Ok ())
  in

  (* this function is failing for untargz *)
  let rec read_rec acc = function
    | 0 -> Tar.return (Ok acc)
    | n ->
        let* content = Tar.read n in
        read_rec (acc ^ content) (n - String.length content)
  in

  (* this one is ok *)
  let read_really n = Tar.really_read n in

  let unzip = Tar_gz.in_gzipped in

  Format.printf "read_really tar...\n";
  (* we don’t care to close fd for this test *)
  let tar = Unix.openfile "test.tar" [ O_RDONLY ] 0 in
  let t = Tar.fold (fold read_really) () in
  let result = Tar_unix.run t tar in
  print_err result;

  Format.printf "read_rec tar...\n";
  let t = Tar.fold (fold (read_rec "")) () in
  let tar = Unix.openfile "test.tar" [ O_RDONLY ] 0 in
  let result = Tar_unix.run t tar in
  print_err result;

  Format.printf "read_really targz...\n";
  let targz = Unix.openfile "test.tar.gz" [ O_RDONLY ] 0 in
  let t = Tar.fold (fold read_really) () in
  let result = Tar_unix.run (unzip t) targz in
  print_err result;

  (* it fails here *)
  Format.printf "read_rec targz...\n";
  let targz = Unix.openfile "test.tar.gz" [ O_RDONLY ] 0 in
  let t = Tar.fold (fold (read_rec "")) () in
  let result = Tar_unix.run (unzip t) targz in
  print_err result;

  ()

[dinosaure]

Sorry, we did not imagine such a use. I've just proposed a patch (see #161) to make your code work.

[dinosaure]

@BChabanne can you confirm that my proposal fix your issue? I will be able to cut a release then 👍

[dinosaure]

So your example works on my side. I hope it works for you. I will merge #161 and cut a release. Thanks for your report - this issue will remains open until you decide to close it 👍.