Re: runtest gdb testsuite oddities.

[Phil Muldoon <pmuldoon at redhat dot com>]

Joel Brobecker <brobecker@adacore.com> writes:

>> Toms check-in today regarding the failure in py-infthread.exp had me
>> exploring why I did not see it.  For some reason when one runs: runtest
>> gdb.python/*.exp the testsuite seems to pass just fine.  But when one
>> runs just one (failing) exp file, the test fails as it should.  I'm not
>> sure why, and my fu regarding dejagnu is very slim.  Can anyone explain
>> what is going on here?  Hopefully it is something silly I have done. 
>
> Each testcase (.exp) is actually part of the same run.  When you think
> of scripts being executed, they are actually files being sourced.
> So what I am thinking happened to you is that, in your normal context,
> you first "ran" (in other words: sourced) a gdb.python testcase that
> called "load_lib gdb-python.exp", and thus defined the
> gdb_py_test_silent_cmd procedure.  "Thanks" to that, by the time dejagnu
> "ran" py-infthread.exp, tcl/expect could find gdb_py_test_silent_cmd,
> even though the "script" is missing a call to load_lib.

It also seems to happen with just a plain old "make check" too (I guess
as it equates to runtest $$/*.exp in each directory.)  The import being
sourced from a previous test unit sounds bogus. (Not that I know how to
fix it, or even if we can.)

Cheers

Phil