Odd test failures - weird.exp, mi-cli.exp

Odd test failures - weird.exp, mi-cli.exp

[Daniel Jacobowitz]

Three more failures for i686-pc-linux-gnu have appeared in the last few
days on my test system.

FAIL: gdb.mi/mi-cli.exp: -interpreter-exec console "set $pc=0x0"
FAIL: gdb.mi/mi2-cli.exp: -interpreter-exec console "set $pc=0x0"
FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o

It's not completely clear whether they are gdb bugs or test bugs.
Here's the log entries:

888-interpreter-exec console "set $pc=0x0"
&"Previous frame inner to this frame (corrupt stack?)\n"
888^error,msg="Previous frame inner to this frame (corrupt stack?)"
(gdb) 
FAIL: gdb.mi/mi-cli.exp: -interpreter-exec console "set $pc=0x0"

This is at test bug - it's pretty funny, actually.  The error is
correct, but depends on the precise garbage on the stack when we set
the PC to zero - in my case, if LD_LIBRARY_PATH contains the three
items that the in-tree expect-bld.sh wrapper puts there, then we'll get
an error.

What's this trying to test, and is there some better way?

(gdb) file object.o
Reading symbols from
/opt/src/binutils/x86-as/gdb/testsuite/object.o...done.
Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
(gdb) FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o

I can't figure out what causes this.  Presumably it is griping about
the libthread_db message, but GDB 6.1 appears to generate the same
thing, and I haven't been getting this failure in the past.  Does
anyone else see this?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: Odd test failures - weird.exp, mi-cli.exp

[Daniel Jacobowitz]

On Sun, Apr 18, 2004 at 11:11:50PM -0400, Daniel Jacobowitz wrote:
> (gdb) file object.o
> Reading symbols from
> /opt/src/binutils/x86-as/gdb/testsuite/object.o...done.
> Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
> (gdb) FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o
> 
> I can't figure out what causes this.  Presumably it is griping about
> the libthread_db message, but GDB 6.1 appears to generate the same
> thing, and I haven't been getting this failure in the past.  Does
> anyone else see this?

This one appears to be:
(gdb) file object.o
Reading symbols from /opt/src/binutils/x86-as/gdb/testsuite/object.o...done.
(gdb) FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o

vs.

Reading symbols from object.o...done.
Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
(gdb) PASS: gdb.stabs/weird.exp: weirdx.o read without error

The most likely candidate is:

2004-04-15  Roland McGrath  <roland@redhat.com>

        * symfile.c (symbol_file_add_with_addrs_or_offsets): Take ABFD as
        argument instead of NAME.
        (symbol_file_add, reread_separate_symbols): Call symfile_bfd_open
        in call to symbol_file_add_with_addrs_or_offsets.
        (build_addr_info): New function, helper for ...
        (symbol_file_add_from_memory): New function.
        (add_symbol_file_from_memory_command): New function using that.
        (_initialize_symfile): Register it for add-symbol-file-from-memory.
        (pre_add_symbol_hook): Add const to argument type.
        * symfile.h (symbol_file_add_from_memory): Declare it.

I guess this is a testsuite bug.  I'll fix it once we decide what to do
about the fluky mi-cli.exp failures in the previous message.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: Odd test failures - weird.exp, mi-cli.exp

[Mark Kettenis]

   Date: Sun, 18 Apr 2004 23:44:59 -0400
   From: Daniel Jacobowitz <drow@false.org>

   On Sun, Apr 18, 2004 at 11:11:50PM -0400, Daniel Jacobowitz wrote:
   > (gdb) file object.o
   > Reading symbols from
   > /opt/src/binutils/x86-as/gdb/testsuite/object.o...done.
   > Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
   > (gdb) FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o
   > 
   > I can't figure out what causes this.  Presumably it is griping about
   > the libthread_db message, but GDB 6.1 appears to generate the same
   > thing, and I haven't been getting this failure in the past.  Does
   > anyone else see this?

Yes, I'm seeing this on at least i386-unknown-freebsd4.7,
amd64-unknown-openbsd3.5 and sparc-unknown-openbsd3.5.

   This one appears to be:
   (gdb) file object.o
   Reading symbols from /opt/src/binutils/x86-as/gdb/testsuite/object.o...done.
   (gdb) FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o

   vs.

   Reading symbols from object.o...done.
   Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
   (gdb) PASS: gdb.stabs/weird.exp: weirdx.o read without error

Yup, although the libthread_db has nothing to do it, so on FreeBSD and
OpenBSD I'm seeing:

   (gdb) file object.o
   Reading symbols from /opt/src/binutils/x86-as/gdb/testsuite/object.o...done.
   (gdb) FAIL: gdb.stabs/weird.exp: Errors reading weirdx.o

   vs.

   (gdb) file object.o
   Reading symbols from object.o...done.
   (gdb) PASS: gdb.stabs/weird.exp: weirdx.o read without error

   I guess this is a testsuite bug.  I'll fix it once we decide what to do
   about the fluky mi-cli.exp failures in the previous message.

Yes please!

Mark