Bugs in g95 reported but not yet fixed as at 20 Aug 2013

1. Real(10) is available in some Linux systems but g95 treats its Infinity 
as NaN. Test program that prints NaN but ought to print Inf or Infinity:
  program testinf10 
    implicit none 
    real(10) inf10
    character(3):: cinf = 'Inf'
    read(cinf,*) inf10
    print *,inf10
  end program testinf10
       
2. Real(10) scalar constants seem to occupy less storage space than 
real(10) scalar variables. The bug may be in transfer. Test program:
  program testtransfer10
    implicit none
    real(10),parameter:: const = 0
    real(10)          :: variable = const
    integer           :: i(1)
    print *,'transfer(variable, i) =',transfer(variable,i)
    print *,'transfer(const   , i) =',transfer(const   ,i)
  end program testtransfer10

3. As F format output requires at least one digit as well as . this
three-line program should print * but it actually prints a decimal point:
  program testf10
    print "(F1.0)", 0.01
  end program testf10

4. If you use SS in an output format the optional + sign is supposed not
to be written, but g95 still writes it if the real value is +Inf (unless
it's kind 10: see bug 1 above). Test program:
  program testplusinf
    implicit none
    character(4):: cinf = '+Inf'
    real           rinf
    read(cinf,*)   rinf
    print "(SS,A,F0.0)",'Inf with SS format = ',rinf 
  end program testplusinf

5. This program is supposed to convert an integer array to quadruple
precision real. On a NetBSD system it prints 1. 3. as it should, but on 
an x86_64 system (on which g95 allows quad precision) it crashes with a 
run-time segmentation fault:
  program testquad
    print *,real((/1,3/),selected_real_kind(30))
  end program testquad

Bugs found later:

6. The arrays a,b should be the same when printed. RHS of an assignment
statement should be completely evaluated before it's substituted into LHS.
  program arraybug
    integer:: a(2) = (/1,2/),b(2)
    b = (/a(2),a(1)+a(2)/)
    a = (/a(2),a(1)+a(2)/)
    print *,a,'should be the same as'
    print *,b,merge('and is.    ','but is not.',maxval(abs(a-b))==0) 
  end program arraybug