While going through xgen22 I have noted the following problems(queries)


  1) The meaning of (say) Indregmodval  is  not  immediately  obvious.    Better
     commenting here would be appreciated.

  2) You accept -1 as invalid astackaddr  meaning  the  local  pointer  invalid.
     Other Ecode machines use 0 here.

  3) You assume that my local pointer from which you  grab  temporaies  contains
     the params but does not allow for the red tape.   This is arbitary ( Imp it
     contians both).   Suggest two further bits allocate in Props to say what is
     included in this.

  4) Your claiming of stackframe space in eprocend does not allow for mode  than
     12 bits of space.   This does occurr from time to time in programs and will
     cause a very hairy bug.   At least abort here.   I deal with this  case  by
     addint the next highest multiple of 4k. Wastes some space but does at least
     work.

  5) You do not believe the local size at  eprocend  using  the  next  temporary
     address if larger.  Something is seriously wrong if this is not right and I
     think you should eabort. This probably put in as a result of confusion

  caused by what is included in this pointer (My para 3)

  6) The smallest code and (apart from the onw word case) fastest way to copy  a
     dislpay is with MVC.

  7) With records by value and strings by value it is not safe  to  assume  that
     the  display  is  within 4096 bytes of r10.   You do assume tyhis both when
     copying it and when using it.

  8) Egenerate object is not in the file(only a duumy needed) but  is  in  GEM's
     file of specs.  This really belongs with the Driver but specs & code should
     be consistent.

  I am cobbling up an IMP ecode generator to investigate fuirther. I will add to
     this list if I find any other problems.