Get (in String; out Item; out Last) in many IO packages for null records is incorrect.

[darkestkhan]

Last is always set to negative value, which under certain circumstances may lead to incorrect results.

-- could quickly write up another, less obscure, example resulting in Constraint_Error
declare     
   High : Integer := 0; -- should be called Low, but whatever
   Last : Integer := 0;     
   Var  : String (1 .. 100);        
   Prep : Preposition_Record;       
   Conj : Interjection_Record;      
   Noun : Noun_Record;      
begin       
   Preposition_Record_IO.Get (Var, Prep, High);     
   Interjection_Record_IO.Get (Var (High .. Var'Last), Conj, Last);     
   High := High + Last; -- High now is equal to High - 1;, thus     
   -- Noun_Record is 'Get' from Var last char of Preposition_Record!        
   Noun_Record_IO.Get (Var (High .. Var'Last), Noun, Last);     
end;        

Last should be set into position of last character in String that was processed inside Get.
In case of Inflections_Package.Interjection_Record_IO (among others) it is set to -1 in all cases (should be 0).
Unfortunately Interjection_Record_IO.Get can't be fixed at the moment due to different bug somewhere between makeinfl.adb and Inflections_Package.Ending_Record_IO.

[mk270]

Thanks for taking the time to create the ticket; I appreciate this duplicates work you'd already done writing this up in a comment in the code.

Is the third argument ("Last") part of an Ada standard/convention, or something created by Whitaker?

[darkestkhan]

@mk270 His various IO packages are based upon Ada.Text_IO. These particular subprograms are based on: http://www.adaic.org/resources/add_content/standards/05rm/html/RM-A-10-1.html paragraph 49; http://www.adaic.org/resources/add_content/standards/05rm/html/RM-A-10-7.html paragraphs 18-20.

[mk270]

Ok, and presumably we should stick to that convention? (and the code currently violates it?)

[darkestkhan]

It would be better to stick to this convention, especially when you consider that calling this Get () subprogram may give you unexpected results. Generally standard Ada subprograms return 0 as position if nothing was processed/found (e.g. Strings.Fixed.Index, Vectors.Find_Index etc.). If these definitions were not based on standard definitions I would ignore it - alas they are, and this can cause confusion.
But as I wrote in issue - it can't be completely fixed atm, due to obscure bug in makeinfl.

[mk270]

Ok, I shall try to sort out makeinfl; the bugs I think are actually in routines that makeinfl calls.

The bad news is that these routines are also called from the src/tools subdirectory, which we've been ignoring.

[darkestkhan]

Now that I think about it - we should have gprfile for src/tools too. I will add one in a moment and do pull request.

[mk270]

Ah good, thanks. There is some documentation somewhere about these tools - I am not sure if they are still necessary/useful, but they may well be.

[darkestkhan]

There is howto file in src/tools. Just checked how build looks for them - many warnings... all over again :(

[mk270]

Now that we have all the callers of get_non_comment_line(), one of which was in tools/ , I have fixed get_non_comment_line() such that it no longer throws exceptions in makeinfl - does this get us any further forward?

[darkestkhan]

Yes, but in different issue (in fact this should fix #36 ). And it does get us further - maybe the entire thing was cause by this. I will check in a moment.