Re: mips.c machine_dependent_reorg fix

[Jeffrey A Law <law at cygnus dot com>]

  In message <200005311512.QAA17020@phal.cygnus.co.uk>you write:
  > In article <10436.959725891@upchuck> you wrote:
  > 
  > : You need to provide a description of why that change is the correct chang
  > e
  > : to make.
  > 
  > 1) It makes it consistent with the CONSTANT_POOL_ADDRESS_P case
True.  But I wasn't sure if that's what was really intended or not.


  > 2b) mode is only used if val is used.
  > 2c) max_internal_pool_size is the maximum total constant pool size that
  >     we're still have to emit, and it shrinks with each constant we examine,
  >     no matter if we do place it in an extra constant pool or if we find
  >     that we don't need to after all.
OK.

  >     The fact that the way this is used incorrectly is IMHO a separate
  >     matter; the most straightforward fix would be to add another variable
  >     that accumulates the size of the currently constructed pool's size,
  >     which is cleared whenever the pool is written out, and taken into
  >     account for the distance computations.
Agreed, we should address this separately.


  > 3) You added the code just two years ago, so you should have a general idea
  >    what it is doing.
I believe it came from the changes to merge mips16 into the mainline tree;
while I have a reasonable idea of what most of the code does, there are 
bits and pieces I do not understand (I'm not the original author, just the
poor bastard that had to pick up the pieces :-)

Go ahead and install the patch.

Thanks,
jeff