This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFC] gdb.server testcases (resend)



Daniel,

I misunderstood the use of gdb.server directory. I thought that this
directory is for testing gdbserver remote debugging functionality. I am
trying to get gdbserver working on ppc. As a starting point I used these
tests. But you are correct. In this perspective my testcases seem useless.

To begin with gdbserver program fails outright on ppc64 PPC970 Altivec
hardware:

 ./gdbserver uranus.ltc.austin.ibm.com:1234 /tmp/server
Process /tmp/server created; pid = 30836
reading register 1: Input/output error
Exiting

I am trying to chase down the problem and get gdbserver working. If you
have any insight into this problem please advice.

Thanks
-----
manjo
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+ Cogito ergo sum                                                          +
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

On Tue, 17 May 2005, Daniel Jacobowitz wrote:

> On Mon, May 16, 2005 at 12:13:17PM -0500, Manoj Iyer wrote:
> >
> > In my previous email I missed the changelog. Here is my complete patch.
> > Please review and comment.
> >
> > 2005-05-16  Manoj Iyer  <manjo@austin.ibm.com>
> >
> >         * gdb.server/server-run.exp: added testcases.
> >         * gdb.server/server.c: added nested function call to test
> >           backtrace.
>
> That's not a changelog; it does not describe what has changed.  Please
> follow the conventions for C code when changing C files in the
> testsuite; each function needs its own entry, for instance.
>
> You've changed the indentation in server.c, away from the GNU style.
> It's not as important to maintain GNU Coding Standards in the
> testsuite, but please don't ignore it without a reason.
>
> > +# test setting a breakpoint
>
> Comments are full sentences, start with capital letters, and end with
> periods.
>
> >  gdb_breakpoint main
> > -gdb_test "continue" "Breakpoint.* main .*" "continue to main"
> > +
> > +gdb_test "continue" ".*Continuing\\..*Breakpoint \[0-9\], main.*at .*$srcfile:\[0-9\]+.*"
> > +
> > +# test if list command displays source code
> > +gdb_test "list" ".*main.*\{.*\}"
> > +
> > +# set breakpoint at a function and test backtrace command
> > +gdb_test "break function3" "Breakpoint 2 at.*file .*$srcfile, line \[0-9\]+."
> > +
> > +gdb_test "continue" ".*Continuing\\..*Breakpoint \[0-9\]+, function3.*at.*$srcfile:\[0-9\]+.*"
>
> You'll need to give names to tests; otherwise this test is going to
> report "PASS: gdb.server/server-run.exp: continue" multiple times.
>
> Manoj, before you revise the patch again, could you explain why you
> want to add these tests?  It is possible to run the entire testsuite
> using gdbserver if you want to specifically validate gdbserver; the
> purpose of the gdb.server directory is:
>
> 1. To make sure that minimal remote protocol support is not
> accidentally broken by people working on native debuggers.
>
> 2. (Someday) to test gdbserver-specific features, like gdbserver
> --attach.
>
> Do these tests add value to #1?  We're already testing that we can
> reach main; on a native configuration that tests breakpoints,
> continuing, and single-stepping (for the dynamic linker breakpoints).
>
> --
> Daniel Jacobowitz
> CodeSourcery, LLC
>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]