remove the cstruct dependency (#137)

remove the cstruct dependency

* Minor refactor
* Use write_(sub)string
Then we don't need to do unsafe conversion to bytes.
* The indice start from an offset but still is absolute. We must take in account this detail into our unmarshal_string implementation
* seek_in expects an absolute value where tar wants to move relatively
from the last position - fix the global_Extended_headers_test
* Fix how we parse parameters (and optimise the loop to be tailcall) from
the usage of String.index_from
* Replace Bytes.create by Bytes.make (and initialise buffer with \000)
* Update Tar_gz to manipulate only string & bytes instead of bistring
* upgrade tar-mirage with the new interface
* Upgrade the implementation of tar-eio
* Tar_gz: Remove unused functions from the READER module type
* Fix Tar_mirage
* tar-mirage: undo some cstruct.t -> bytes conversions
* whitespace cleanup

Co-authored-by: Reynir Björnsson <[email protected]>
Co-authored-by: Calascibetta Romain <[email protected]>
Co-authored-by: Kate <[email protected]>

CHANGES.md

@@ -4,20 +4,20 @@
   (#119, #120, @MisterDA)
 - Support GNU LongLink and LongName. Prior, `Tar.HeaderWriter` and
   `Tar.HeaderReader` supported both, but `Tar.Header.Link` only had `LongLink`
-  and (de)serialized to (from) GNU LongName.
+  and (de)serialized to (from) GNU LongName (#127)
 - Compatibility level when reading / parsing is removed. Only GNU
   LongLink/LongName extensions were affected by the compatibility level when
-  reading.
+  reading (#127)
 - Add module types `Tar.HEADERREADER` and `Tar.HEADERWRITER` describing the
-  output of `Tar.HeaderReader` and `Tar.HeaderWriter` respectively.
-- Types `Tar.READER.t` and `Tar.WRITER.t` are renamed to `io`.
-- Add `write_global` function for writing a global PAX extended header.
-- Rework IO-specific modules (tar-unix etc.) harmonizing them.
+  output of `Tar.HeaderReader` and `Tar.HeaderWriter` respectively (#127)
+- Types `Tar.READER.t` and `Tar.WRITER.t` are renamed to `io` (#127)
+- Add `write_global` function for writing a global PAX extended header (#127)
+- Rework IO-specific modules (tar-unix etc.) harmonizing them (#127)
 - Avoid exceptions in tar and use result instead. The exceptions
-  `End_of_stream` and `Checksum_mismatch` are removed.
-- Remove the `Tar_cstruct` module as it was unused.
-- Remove debug printers.
-- Finally remove the unused camlp-streams dependency.
+  `End_of_stream` and `Checksum_mismatch` are removed (#127)
+- Remove the `Tar_cstruct` module as it was unused (#127)
+- Remove debug printers (#127)
+- Finally remove the unused camlp-streams dependency (#127)
 
 ## v2.6.0 (2023-09-07)
 

bin/otar.ml

@@ -22,22 +22,13 @@ module Tar_gz = Tar_gz.Make
           let return x = x end)
   (struct type out_channel = Stdlib.out_channel
           type 'a io = 'a
-          let really_write oc cs =
-            let str = Cstruct.to_string cs in
+          let really_write oc str =
             output_string oc str end)
   (struct type in_channel = Stdlib.in_channel
           type 'a io = 'a
-          let really_read ic cs =
-            let len = Cstruct.length cs in
-            let buf = Bytes.create len in
-            really_input ic buf 0 len ;
-            Cstruct.blit_from_bytes buf 0 cs 0 len
-          let skip ic len = really_read ic (Cstruct.create len)
-          let read ic cs =
-            let max = Cstruct.length cs in
-            let buf = Bytes.create max in
-            let len = input ic buf 0 max in
-            Cstruct.blit_from_bytes buf 0 cs 0 len ; len end)
+          let read ic buf =
+            input ic buf 0 (Bytes.length buf)
+        end)
 
 
 let ( / ) = Filename.concat
@@ -107,7 +98,7 @@ let bytes_to_size ?(decimals = 2) ppf = function
 
 let list filename =
   let ic = open_in filename in
-  let ic = Tar_gz.of_in_channel ~internal:(Cstruct.create 0x1000) ic in
+  let ic = Tar_gz.of_in_channel ~internal:(De.bigstring_create 0x1000) ic in
   let rec go global () = match Tar_gz.HeaderReader.read ~global ic with
     | Ok (hdr, global) ->
       Format.printf "%s (%s, %a)\n%!"

dune-project

@@ -32,7 +32,6 @@
  (tags ("org:xapi-project" "org:mirage"))
  (depends
   (ocaml (>= 4.08.0))
-  (cstruct (>= 6.0.0))
   (decompress (>= 1.5.1))
  )
 )
@@ -47,8 +46,6 @@
  (tags ("org:xapi-project" "org:mirage"))
  (depends
   (ocaml (>= 4.08.0))
-  (cstruct (>= 6.0.0))
-  cstruct-lwt
   lwt
   (tar (= :version))
  )
@@ -66,7 +63,7 @@
  (conflicts (result (< 1.5)))
  (depends
   (ocaml (>= 4.08.0))
-  (cstruct (>= 1.9.0))
+  (cstruct (>= 6.0.0))
   (lwt (>= 5.6.0))
   (mirage-block (>= 2.0.0))
   (mirage-clock (>= 4.0.0))

eio/tar_eio.ml

@@ -27,7 +27,11 @@ end
 module Io = struct
   type in_channel = Flow.source
   type 'a io = 'a
-  let really_read f b = Flow.read_exact f b
+  let really_read f b =
+    let len = Bytes.length b in
+    let cs = Cstruct.create len in
+    Flow.read_exact f cs;
+    Cstruct.blit_to_bytes cs 0 b 0 len
   let skip f (n: int) =
     let buffer_size = 32768 in
     let buffer = Cstruct.create buffer_size in
@@ -36,15 +40,18 @@ module Io = struct
       else
         let amount = min n buffer_size in
         let block = Cstruct.sub buffer 0 amount in
-        really_read f block;
+        Flow.read_exact f block;
         loop (n - amount) in
     loop n
 
   type out_channel = Flow.sink
-  let really_write f b = Flow.write f [ b ]
+  let really_write f str = Flow.write f [ Cstruct.of_string str ]
 end
 
-include Io
+let really_read = Flow.read_exact
+let skip = Io.skip
+let really_write f b = Flow.write f [ b ]
+
 module HeaderReader = Tar.HeaderReader(Monad)(Io)
 module HeaderWriter = Tar.HeaderWriter(Monad)(Io)
 

lib/dune

@@ -2,8 +2,7 @@
  (name tar)
  (modules tar)
  (public_name tar)
- (wrapped false)
- (libraries cstruct))
+ (wrapped false))
 
 (library
  (name tar_gz)