Re: GDB 5.0 testsuite failures

Re: GDB 5.0 testsuite failures

[Michael Elizabeth Chastain]

Hi Matthew,

Thanks for posting this.  I appreciate seeing gdb test results from diverse
architectures.

Summary: this gdb looks healthy to me.   There is one major bug:
if the user uses gdb to call a function in the target program, and that
function has floating point parameters or a floating point return value,
the call will not work.

There are probably some more minor bugs but I can't tell from the gdb.sum
file.

BTW, please note that the test run produces a "gdb.sum" file and a "gdb.log"
file.  The "gdb.log" file has a lot more information and it's a better file
to put up for ftp, rather than capturing the output of "make check".

Michael Elizabeth Chastain
<chastain@redhat.com>
"love without fear"

  ERROR: 26

    23 errors in gdb.stabs/weird.exp
      These look harmless; your assembler does not like .stabs directives.

    3 errors from 'make' at the end of the run
      More fluff.

  WARNING: 3

    1 warning in gdb.base/corefile.exp
      'can't generate a core file' -- does not look like a gdb problem

    2 warnings in gdb.stabs/weird.exp
      Yeah, we know it doesn't like stabs.

  FAIL: 224

    190 fails in gdb.stabs/weird.exp
      Will this test case shut up already?!

    7 fails in gdb.base/callfwmall.exp
      These are all related to calling functions with floating point
      parameters or floating point return values.  This looks like a
      real bug in gdb.

    7 fails in gdb.base/callfuncs.exp
      These are all related to calling functions with floating point
      parameters or floating point return values.  This looks like a
      real bug in gdb.

    5 fails in gdb.c++/ovldbreak.exp
      I would need to see gdb.log to tell whether this is a real bug or not.

    3 fails in gdb.threads/pthreads.exp
      Compiler can't find pthreads library so all these tests fail.

    2 fails in gdb.base/commands.exp
      One of these is a known bug in the test script (gnats gdb/14).
      The other looks like a bug in gdb.

    2 fails in gdb.base/break.exp
      I don't know anything about these failures.

    1 fail  in gdb.base/varargs.exp
      This is another call-to-function with floating point parameters.

    1 fail  in gdb.base/signals.exp
      This is marked as 'known SV4 bug'.  It may mean that 'next' will
      not work on your machine and we can't fix it.

    1 fail  in gdb.base/selftest.exp
      I don't know anything about this failure.

    1 fail  in gdb.base/interrupt.exp
      I don't know anything about this failure.

    1 fail  in gdb.base/annota1.exp
      I don't know anything about this failure.

Re: GDB 5.0 testsuite failures

[Matt Schalit]

Michael Elizabeth Chastain wrote:
> 
> Hi Matthew,
> 
> Thanks for posting this.  I appreciate seeing gdb test results from diverse
> architectures.


Thanks for the quick reply.  After this, is it proper to
post a gbd backtrace here and ask for help?  
It's only 11 lines.

 
> Summary: this gdb looks healthy to me.   There is one major bug:
> if the user uses gdb to call a function in the target program, and that
> function has floating point parameters or a floating point return value,
> the call will not work.




If that bug doesn't occur when I do the following,
then I'm safe:

 (gdb) file /usr/local/bin/gnomecc
 (gdb) run
 (gdb) bt





> There are probably some more minor bugs but I can't tell from the gdb.sum
> file.
> 
> BTW, please note that the test run produces a "gdb.sum" file and a "gdb.log"
> file.  The "gdb.log" file has a lot more information and it's a better file
> to put up for ftp, rather than capturing the output of "make check".


  ftp://63.194.213.179:710/pub/Random/gdb.log



 
> Michael Elizabeth Chastain
> <chastain@redhat.com>
> "love without fear"
> 
>   ERROR: 26
> 
>     23 errors in gdb.stabs/weird.exp
>       These look harmless; your assembler does not like .stabs directives.
> 
>     3 errors from 'make' at the end of the run
>       More fluff.
> 
>   WARNING: 3
> 
>     1 warning in gdb.base/corefile.exp
>       'can't generate a core file' -- does not look like a gdb problem
> 
>     2 warnings in gdb.stabs/weird.exp
>       Yeah, we know it doesn't like stabs.
> 
>   FAIL: 224
> 
>     190 fails in gdb.stabs/weird.exp
>       Will this test case shut up already?!
> 
>     7 fails in gdb.base/callfwmall.exp
>       These are all related to calling functions with floating point
>       parameters or floating point return values.  This looks like a
>       real bug in gdb.
> 
>     7 fails in gdb.base/callfuncs.exp
>       These are all related to calling functions with floating point
>       parameters or floating point return values.  This looks like a
>       real bug in gdb.
> 
>     5 fails in gdb.c++/ovldbreak.exp
>       I would need to see gdb.log to tell whether this is a real bug or not.
> 
>     3 fails in gdb.threads/pthreads.exp
>       Compiler can't find pthreads library so all these tests fail.



When I throw -pthreads as a LIBS during the build of an app,
it gets added to all the other -lm -lgen -socket -lnsl libs.
But that's the special compiler flag for gcc.  I do have these:

/usr/lib/libthread.so.1
/usr/lib/libthreadT.so.1
/usr/lib/libthread.so
/usr/lib/libthreadT.so





 
>     2 fails in gdb.base/commands.exp
>       One of these is a known bug in the test script (gnats gdb/14).
>       The other looks like a bug in gdb.
> 
>     2 fails in gdb.base/break.exp
>       I don't know anything about these failures.
> 
>     1 fail  in gdb.base/varargs.exp
>       This is another call-to-function with floating point parameters.
> 
>     1 fail  in gdb.base/signals.exp
>       This is marked as 'known SV4 bug'.  It may mean that 'next' will
>       not work on your machine and we can't fix it.
> 
>     1 fail  in gdb.base/selftest.exp
>       I don't know anything about this failure.
> 
>     1 fail  in gdb.base/interrupt.exp
>       I don't know anything about this failure.
> 
>     1 fail  in gdb.base/annota1.exp
>       I don't know anything about this failure.


Ok.  Thanks again,
Matthew

Re: GDB 5.0 testsuite failures

[Michael Elizabeth Chastain]

Hi Matthew,

> Thanks for the quick reply.  After this, is it proper to
> post a gbd backtrace here and ask for help?  
> It's only 11 lines.

Sure, go ahead.  The availability of help varies widely from hour to
hour though.

> If that bug doesn't occur when I do the following,
> then I'm safe:
> 
>  (gdb) file /usr/local/bin/gnomecc
>  (gdb) run
>  (gdb) bt

Yeah, your gdb ought to be able to backtrace just fine.

> ftp://63.194.213.179:710/pub/Random/gdb.log

Would it be OK with you if I entered this in the public test report
database at http://sources.redhat.com/cgi-bin/gdb.cgi ?

Michael Elizabeth Chastain
<chastain@redhat.com>
"love without fear"

Re: GDB 5.0 testsuite failures

[Matt Schalit]

Michael Elizabeth Chastain wrote:
> 
> Hi Matthew,


> > ftp://63.194.213.179:710/pub/Random/gdb.log
> 
> Would it be OK with you if I entered this in the public test report
> database at http://sources.redhat.com/cgi-bin/gdb.cgi ?


Sure, that's no problem.  I'm going to take it off the
ftp site in a week, so I figure you'll either paste it
into the database or grab a copy for yourself.

Thanks,
Matt

GDB 5.0 testsuite failures

[Matt Schalit]

Hi Folks,

I had  224 unexpected testsuite failures with gdb-5.0,
which I compiled and ran on an i586-sco-sysv5uw7.1.1,
which is UnixWare 7.1.1.

I copied the ouput of  gmake check to a file so that
you can look at it:

      ftp://63.194.213.179:710/pub/Random/gdb.check.log


And here's what I used to compile 5.0 with gcc 2.95.2.1,
and gnu-make-3.79.1.

  LIBS="-lsocket -lnsl"  ./configure --enable-shared --disable-nls 
  --x-includes=/usr/X/include --x-libraries=/usr/X/lib



Can I assume that the gdb is ok?
Thanks,
Matthew