Re: [pdf-devel] [PATCH] Error management in type4 functions

[jemarch]

Hi David.

Regarding your patch on error management in type4 functions...

  - Fix a bug in the real numbers reader. I wrote parse_real function which
    works, though it does not check overflows. It should be moved to the
    tokenizer module.

Superb, thanks.

  - A code as the following compiled without errors, but the evaluator crashed:
      {
        ...
        {

        ...
           }
             ...
           } if
        ...
        } if
      }
      The patch fixes this bug.

  - pdf_fp_func_4_new reports errors correctly. It reports the position
    of the error in the code too. I use the return value to do this.

  - pdf_fp_func_eval reports interpretation errors in type 4 functions
    correctly. I added some debug information to the function structure
    in addition to the opcode. So, the evaluator can get the offset of a
    interpretation error in the code, and inspect the current context
    (the stack). However, the documentation does not describe the debug
    structure yet.

Quite useful. I just created a task covering the documentation of the
new facility.

  - Add negative unit testing for the above-named.


  Finally the type 4 evaluator does not check some `bad type'
  errors. The reason is that two special real numbers stand for logic
  values:

     #define REP_TRUE   (1.202056903159594285399738162)
     #define REP_FALSE  (-15116315767.09215686274509804)

  It is not very smart for me reporting errors so.
  I wait you give me some feedback.

What `bad type' errors are those?

In the meanwhile I applied the patch in the trunk. Many thanks :)

-- 
Jose E. Marchesi  <address@hidden>
                  http://www.jemarch.net
GNU Project       http://www.gnu.org