A bug in mark_constants () in varasm.c

[Jeffrey A Law]

  In message < 199803310212.EAA02501@quatramaran.ens.fr >you write:
  > The main function is roughly 1000 lines long. I promptly gave up.
  > 
  > main begins with stuff such as raising limits on stack size, or
  > finding out a program name, including a VMS patch...
I don't consider cccp.c one of the better written modules in the
compiler.  In fact it's one of the worst.

  > What is unsurprising at this point is that the part that raises
  > limits on stack size is duplicated in genattrtab.c and toplev.c.
Not suprising.

  > I don't understand that. There must have been someone who wrote this
  > code in at some point.   Why is there three instances of the code ?
They didn't take the time to break it out into a file that could be
linked into the different programs where such code would be useful
(the three modules you mention are all part of separate programs).

  > I would like to comment that commonality of code and function would be
  > much easier to see with less obfuscation.
Can't argue with that.  All I can say is we're addressing such problems
as quickly as we can -- see the system.h cutover.  I'm not sure what
the next big "make everyone's life easier" change will be, but system.h
certainly won't be the last!

jeff