bad looking numbers for gdb 5.3.91

[Michael Elizabeth Chastain]

Yesterday I wrote:
> The problem is that gdb.base/fileio.exp bit the mighty twinkie
> with gcc HEAD, both dwarf-2 and gstabs+.  I don't know whose problem
> this is.

Progress report: this looks like a regression bug in gcc HEAD,
sometime between 2003-09-03 00:00:00 UTC and 2003-09-04 00:00:00 UTC.

Last report:

 (gdb) break main
 Breakpoint 1: file /berman/fsf/_today_/source/gdb/gdb_6_0-branch/src/gdb/testsuite/gdb.base/fileio.c, line 490.

This report:

 (gdb) break main
 Breakpoint 1: file stat.h, line 490.

gcc is sticking bad ".file stat.h ..." lines into the generated
assembly code.  It's about 95% certain that this is a gcc bug,
not a gdb bug (always a 5% chance that I am making a "Doh!" kind
of mistake reading something).

It's interesting that fileio.exp is the only test that catches this.
We could use more tests that run in a hosted posix environment
and use the standard libraries and stuff.

I'm still working on isolating this.  Normally I wouldn't take up
bandwidth talking about this before I completely isolated the bug, but
it freaks me when gdb_6_0-branch has a big regression in test results so
late in the release cycle.

Michael C