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: 3.0.1 Freeze




--On Tuesday, August 07, 2001 01:02:12 AM +0200 Gabriel Dos_Reis 
<gdosreis@sophia.inria.fr> wrote:

> Before we went down the "stagging hearders" road, Gerald could build the
> GCC-3.0.x on solaris-2.8.  Now he can't because of the multilib issue
> which is showing on more targets -- David already complained about
> that for GCC-3.0 vs. AIX.
>
> | What are the plans for dealing with them?  When do we expect these
> | plans to be implemented?
>
> The plan we agreed on was to have <limits> pregenerated.  That means
> we need to

OK, I remember this issue.

It is not good that the staging headers went in on the branch without
solving this problem and this close to a release.  Let's avoid that
in the future.

Here's the plan:

  1. Build a generic std_limits.h that has a structure like:

     #ifndef __glibcpp_int_bits
     #define __glibcpp_int_bits 32
     #endif

     #if __glibcpp_int_bits == 32
     #ifndef __glibcpp_int_max
     #define __glibcpp_int_max <right value>
     #endif
     #ifndef __glibcpp_long_max
     #define __glibcpp_long_max <right value>
     #endif
     ...
     #elif __glibcpp_int_bits == 64
     ...
     #endif

     <stuff involving __glibcpp_long_max>

  2. Go through the V3 OS configs and set the
     __glibcpp_target_int_bits, etc., macros as best as you
     can.  Ask target maintainers for help with parameters
     you don't know or for machines to which you don't have
     access.  Ask people to run mknumeric_limits and send you
     the output.  People, please help Gaby with this.

  3. Change the build structure to make this work.

This is now your top priority for CodeSourcery, as part of our LANL
work.

Do you have time to do this in the next 2-3 days?  If not, please
let me know ASAP and I will do some of the bits.  I suggest that you
start with the code bits; it will be easier for me to get the
configury right than stuff about denormalized floating point
numbers. :-)

If we cannot solve these problems quickly, we will have to revert the
staging headers patches -- even though we all agree they are an
improvement.

Thank you,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com


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