🚀 go-pugleaf

I'm running gdb 3.5 under 386/ix 2.0.2, compiling everything with gcc 3.7
and using the vendor's assembler and linker, not COFF encapsulation.

In an object file, the introductory symbol entry for an enum, which usually
has a name like .2fake, includes an auxiliary entry with the symbol index one
past the end of the .eos entry at the end of the member list.  There seems to
be a bug in the 386/ix linker that messes up this index now and then.  GDB
believes the index and gets out of sync reading the symbol table, causing
various error reports, usually .bb/.eb mismatch.  Since every enum group ends
with a .eos, this patch adds defensivee code to make it stop reading enum
members when it sees the .eos even if the index hasn't been reached.

I note that the routine that reads structure entries, which are arranged the
same way, already has similar defensive code, so I presume that someone ran
into the same bug before in structure symbols.

The bug, by the way, is definitely in the vendor's linker, not the assembler.
I stared at some huge hex object file dumps to be sure.  I know that in a
perfect world I'd use gas and gld, maybe later.

*** gdb-3.5/coffread.c	Tue Mar 20 17:11:23 1990
--- gdb-watch/coffread.c	Tue Mar 27 14:10:56 1990
***************
*** 1907,1912 ****
--- 1907,1913 ----
    register struct symbol *sym;
    register struct type *type;
    int nsyms = 0;
+   int done = 0;
    struct pending **symlist;
    struct coff_symbol member_sym;
    register struct coff_symbol *ms = &member_sym;
***************
*** 1928,1934 ****
      symlist = &file_symbols;
    osyms = *symlist;
  
!   while (symnum < lastsym && symnum < nlist_nsyms_global)
      {
        read_one_sym (ms, &sub_sym, &sub_aux);
        name = ms->c_name;
--- 1929,1935 ----
      symlist = &file_symbols;
    osyms = *symlist;
  
!   while (!done && symnum < lastsym && symnum < nlist_nsyms_global)
      {
        read_one_sym (ms, &sub_sym, &sub_aux);
        name = ms->c_name;
***************
*** 1949,1954 ****
--- 1950,1956 ----
  	    break;
  
  	  case C_EOS:
+ 	    done = 1;	/* in case the count was wrong */
  	    break;
  	}
      }
-- 
John R. Levine, Segue Software, POB 349, Cambridge MA 02238, +1 617 864 9650
johnl@esegue.segue.boston.ma.us, {ima|lotus|spdcc}!esegue!johnl
"Now, we are all jelly doughnuts."