GCC 3.5 and debug info generation

[Daniel Jacobowitz]

On Wed, Aug 25, 2004 at 02:18:06PM -0700, Devang Patel wrote:
> This is just heads up.
> 
> In last six month or so, quality of debugging info. generated by GCC  
> 3.5 has regressed significantly (as per GDB test suite results).
> 
> According to Michael Chastain's test results
> http://gcc.gnu.org/ml/gcc-testresults/2004-08/msg00833.html
> http://www.shout.net/~mec/sunday/current/Non-PASS-gdb-6.2-suite- 
> HEAD.html
> 
> Use of GCC-3.5 results in 925 and 905 failures for STABS and DWARFS  
> format respectively. For the same version of GDB, however use of  
> GCC-3_4-branch results in 33 and 55 failures.
> 
> Above numbers are for i686-pc-linux-gnu. We are seeing similar (not  
> identical) numbers for powerpc-darwin-apple using different GDB  
> version. Right now we are seeing 859 failures for STABS, where as we  
> saw only 239 failures in mid March.

Have you characterized the failures?  Michael has been fixing them;
most of them were caused by the 'implicit definition of built-in
function abort' warning.  He's been filing GCC bugs as he notices
problems, and last I checked most of them were fixed.

-- 
Daniel Jacobowitz