This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: aix EXIT_SUCCESS/EXIT_FAILURE stuff




  In message <378B84C6.5B87AD7F@hermes.si>you write:
  > Found it! Unless i'm completely off my rocker ...
  > 
  > The culprit is definitely fix-header. The problem shows up when the
  > native compiler is used for the bootstrap. Here's what happens:
  > 
  > fix-header.c defines
  > 
  > /* Special name to indicate a continuation line in std_include_table.  */
  > const char CONTINUED[] = "";
  > 
  > `CONTINUED' is used in `std_include_table' to indicate entry
  > continuation. However, inside `main', in the loop which searches
  > `std_include_table':
  > 
[ ... ]

  > (gdb) p CONTINUED
  > $14 = 0x2000d094
  > (gdb) p std_include_table
  > $15 = {{name = 0x2000a314, flags = 1, names = 0x2000c1a4},
  >        {name = 0x2000a31c, flags = 0, names = 0x2000c210},
  >        {name = 0x2000a328, flags = 513, names = 0x2000a330},
  >        {name = 0x2000a338, flags = 1, names = 0x2000c234},
  >        {name = 0x2000a344, flags = 0, names = 0x2000a34c},
  >        {name = 0x2000a360, flags = 0, names = 0x2000a368},
  >        {name = 0x2000a37c, flags = 1, names = 0x2000a388},
  >        {name = 0x2000a3a0, flags = 1, names = 0x2000c33c},
  > here-> {name = 0x2000a310, flags = 513, names = 0x2000a3a8},
  >        {name = 0x2000a3b4, flags = 0, names = 0x2000a3bc},
  >        {name = 0x2000a3d0, flags = 1, names = 0x2000a3dc},
  >        {name = 0x2000a3ec, flags = 1, names = 0x2000a3f8},
  >        {name = 0x2000a404, flags = 1, names = 0x2000c3ac},
  > here-> {name = 0x2000a310, flags = 0, names = 0x2000a40c},
  > here-> {name = 0x2000a310, flags = 0, names = 0x2000a41c},
[ ... ]

  > Notice that the values in the table are exactly different than
  > `CONTINUED' ... there seems to be an "optimisation" going on here,
  > possibly because `CONTINUED' is a const array. I have no idea
  > Whether this optimisation is valid or not. The only definite thing
  > I found is that the string is duplicated at different addresses. Yuck.
Interesting.  I do not see this behavior on my aix4.2 box.  I've got a build
running on a 4.1.3 box, but it will not finish for a while.

What options are you passing to the aix compiler?

David -- I know you've seen this behavior (this is the 
EXIT_SUCCESS/EXIT_FAILURE
stuff).  Can you see if the contents of std_include_table are busted in a
manner similar to what Branko has shown above?  

jeff


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]