This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: PATCH: tests for MI commands
Nick Roberts wrote:
> Given the scarce resources for this unpaid work I ask you again to
> consider removing the mi2-*.exp files and just test the current version
> of MI. The reality is that they serve no useful purpose (mi-*.exp tests
> would still pass if GDB was invoked with "-i=mi2") and they require
> effort that could be better spent on other things.
>
> We can discuss this *after* you either:
> a) Fixed the test failures.
>
> b) Backed out your patch.
I'd like to do a) by removing the mi2-*.exp files ;-). There doesn't seem
much point in making a load of changes if we are going to just remove the
files afterwards.
> You promised not to do this again, and yet you did.
Let's remember that the testsuite is there to prevent regressions in GDB and
not as an end in itself. Presumably the focus should be on having the right
tests in the suite and not just on ensuring that everything passes.
Also remember that you're not the only person working on GDB. :-)
If I'm testing my patch and a bunch of MI tests start failing all
of a sudden, how do I tell the difference between some terrible
bug that I introduced and somebody else leaving some tests broken?
It wastes everybody's time as they try to figure out whether they're
at fault for the breakage. So the rule is that every change should
leave the testsuite results in the same state or better, and yes,
sometimes it means fixing things that you're pretty certain are
going away soon.
The crafty strategize this by submitting the patch to delete the
unwanted tests first, then voila, the code patch just happens
not to cause any visible regressions!
Stan