Re: [RFA] new test for separate debug info

[mec.gnu@mindspring.com (Michael Elizabeth Chastain)]

My test bed says:

  binutils 2.13.90.0.2-2-rh, dwarf-2
  binutils 2.13.90.0.2-2-rh, stabs+
  binutils 2.14, dwarf-2
  binutils 2.14, stabs+
    1 UNSUPPORTED
    # okay

  binutils HEAD, dwarf-2
    77 PASS
    # okay

  binutils HEAD, stabs+
    3 ERROR, 7 PASS, 44 FAIL, 1 UNRESOLVED
    # not okay

The results are the same for all versions of gcc tested.

Here is an excerpt from gdb.log:
 
  # gdb HEAD, gcc 3.3.2, binutils HEAD, -gstabs+
  (gdb) dir /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base
  Source directories searched: /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base:$cdir:$cwd
  (gdb) file /berman/fsf/_today_/berman/test/gdb.base/sepdebug
  Reading symbols from /berman/fsf/_today_/berman/test/gdb.base/sepdebug...Reading symbols from /berman/fsf/_today_/berman/test/gdb.base/.debug/sepdebug.debug...Can't find any code sections in symbol file
  (gdb) ERROR: couldn't load /berman/fsf/_today_/berman/test/gdb.base/sepdebug into gdb.
  delete breakpoints
  (gdb) UNRESOLVED: gdb.base/sepdebug.exp: Delete all breakpoints when none
  From gdb-patches-return-28583-mec.gnu=mindspring.com@sources.redhat.com  Wed Nov 12 17:26:42 2003

It looks like either binutils or gdb needs some work.

If you could kludge the test to handle "Can't find any code sections
in symbol file" then I would be happy with the test script as a test
script.

I have put up a test result tarball:

  ftp://ftp.shout.net/pub/users/mec/gdb/2003-11-10-sepdebug.tar.gz

The interesting files are:

  native,HEAD,3.3.2,HEAD,stabs+,2/
    gdb.log
    test.tar.gz

I actually run the tests with:

  runtest ...  --target_board unix/gdb:debug_flags=-gdwarf-2\\\\\\ -g2 ...
  runtest ...  --target_board unix/gdb:debug_flags=-gstabs+\\\\\\ -g2 ...

It might be some weird option-processing problem, but I doubt it,
because I run dwarf-2 and stabs+ in the same way.  That is, I don't
just run dwarf-2 by leaving the debug_flags as default.
    
Michael C