jon.recoil.org

Module Misc.Magic_number

a typical magic number is "Caml1999I011"; it is formed of an alphanumeric prefix, here Caml1990I, followed by a version, here 011. The prefix identifies the kind of the versioned data: here the I indicates that it is the magic number for .cmi files.

All magic numbers have the same byte length, magic_length, and this is important for users as it gives them the number of bytes to read to obtain the byte sequence that should be a magic number. Typical user code will look like:

  let ic = open_in_bin path in
  let magic =
    try really_input_string ic Magic_number.magic_length
    with End_of_file -> ... in
  match Magic_number.parse magic with
  | Error parse_error -> ...
  | Ok info -> ...

A given compiler version expects one specific version for each kind of object file, and will fail if given an unsupported version. Because versions grow monotonically, you can compare the parsed version with the expected "current version" for a kind, to tell whether the wrong-magic object file comes from the past or from the future.

An example of code block that expects the "currently supported version" of a given kind of magic numbers, here Cmxa, is as follows:

  let ic = open_in_bin path in
  begin
    try Magic_number.(expect_current Cmxa (get_info ic)) with
    | Parse_error error -> ...
    | Unexpected error -> ...
  end;
  ...

Parse errors distinguish inputs that are Not_a_magic_number str, which are likely to come from the file being completely different, and Truncated str, raised by headers that are the (possibly empty) prefix of a valid magic number.

Unexpected errors correspond to valid magic numbers that are not the one expected, either because it corresponds to a different kind, or to a newer or older version.

The helper functions explain_parse_error and explain_unexpected_error will generate a textual explanation of each error, for use in error messages.

type version = int
type kind =
  1. | Exec
  2. | Cmi
  3. | Cmo
  4. | Cma
  5. | Cmx
  6. | Cmxa
  7. | Cmxs
  8. | Cmt
  9. | Cms
  10. | Ast_impl
  11. | Ast_intf
type info = {
  1. kind : Misc.Magic_number.kind;
  2. version : Misc.Magic_number.version;
    (*

    Note: some versions of the compiler use the same version suffix for all kinds, but others use different versions counters for different kinds. We may only assume that versions are growing monotonically (not necessarily always by one) between compiler versions.

    *)
}
type raw = string

the type of raw magic numbers, such as "Caml1999A027" for the .cma files of OCaml 4.10

Parsing magic numbers

type parse_error =
  1. | Truncated of string
  2. | Not_a_magic_number of string
val explain_parse_error : Misc.Magic_number.kind option -> Misc.Magic_number.parse_error -> string

Produces an explanation for a parse error. If no kind is provided, we use an unspecific formulation suggesting that any compiler-produced object file would have been satisfying.

Parses a raw magic number

Read a raw magic number from an input channel.

If the data read str is not a valid magic number, it can be recovered from the Truncated str | Not_a_magic_number str payload of the Error parse_error case.

If parsing succeeds with an Ok info result, we know that exactly magic_length bytes have been consumed from the input_channel.

If you also wish to enforce that the magic number is at the current version, see read_current_info below.

val magic_length : int

all magic numbers take the same number of bytes

Checking that magic numbers are current

type 'a unexpected = {
  1. expected : 'a;
  2. actual : 'a;
}

check_current kind info checks that the provided magic info is the current version of kind's magic header.

val explain_unexpected_error : Misc.Magic_number.unexpected_error -> string

Provides an explanation of the unexpected_error.

type error =
  1. | Parse_error of Misc.Magic_number.parse_error
  2. | Unexpected_error of Misc.Magic_number.unexpected_error

Read a magic number as read_info, and check that it is the current version as its kind. If the expected_kind argument is None, any kind is accepted.

Information on magic numbers

val string_of_kind : Misc.Magic_number.kind -> string

a user-printable string for a kind, eg. "exec" or "cmo", to use in error messages.

val human_name_of_kind : Misc.Magic_number.kind -> string

a user-meaningful name for a kind, eg. "executable file" or "bytecode object file", to use in error messages.

the current magic number of each kind

the current version of each kind

Raw representations

Mainly for internal usage and testing.

type raw_kind = string

the type of raw magic numbers kinds, such as "Caml1999A" for .cma files

parse a raw kind into a kind

the current raw representation of a kind.

In some cases the raw representation of a kind has changed over compiler versions, so other files of the same kind may have different raw kinds. Note that all currently known cases are parsed correctly by parse_kind.

A valid raw representation of the magic number.

Due to past and future changes in the string representation of magic numbers, we cannot guarantee that the raw strings returned for past and future versions actually match the expectations of those compilers. The representation is accurate for current versions, and it is correctly parsed back into the desired version by the parsing functions above.

val all_kinds : Misc.Magic_number.kind list