Re: [RFA]: Fix for pending breakpoints in manually loaded/unloaded shlibs

[Jeff Johnston <jjohnstn at redhat dot com>]

Michael Chastain wrote:
> Jeff Johnston <jjohnstn@redhat.com> wrote:
>
> > The test should behave the same because the gdb code that generates
> > the message it is checking hasn't been committed yet.  On retrospect,
> > perhaps I should not have committed the testcase in with this
> > particular check in place (the test itself is a valid one regardless).
> > There has been some questions regarding whether I should be using an
> > observer or not.  Knowing that, do I still need to do the following or
> > can I check the change I attached in once I get final approval on the
> > code and most importantly, the message to be issued?
>
>
> I don't quite follow you, but if you are asking: can you check in
> a change to the test script before you check in a change to gdb:
> in general, you can do that.
>
> If the test script accepts both old+new messages, and the new message is
> not wildly more complex than the old message, then testing with the old
> message alone is good enough for getting the test script approved.
> Just pop out the new patch and say how / what system you tested on.
>
> If that's not what you mean, I'm confused.

The test fails now because gdb is not issuing the message that is currently in 
the test script.  The message in the test script is a new message which does not 
currently exist and never did in any gdb release.  When I get approval to check 
the gdb code in "and" change the test script, the test will succeed because it 
will match the actual message checked in.  I don't see a version issue of old 
gdb looked for x and new gdb looks for y.  Old gdb will always fail and new gdb 
will always work.  Thus, I was wondering if I still need to have checks for old 
and new messages in the script.

-- Jeff J.