This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: iconvdata regression
On Sat, Mar 31, 2012 at 6:27 PM, David Miller <davem@davemloft.net> wrote:
>> Do we agree that build breakage is *much* more serious since it
>> impacts reg testing?
>
> This iconvdata regression one effected reg testing too, I don't see
> what the difference might be.
Yes, but only if your testing for "No failures."
> A build regression introducing change could potentially just as easily
> be isolated and reverted in the developer's local sources, right?
No. A build regression prevents automated reg-testing from being able
to *run* the reg test since building is a prerequisite. You go from
having regions of commits where the test passes, and fails, to having
a third state of unknown when the build doesn't complete. The third
state of unknown is a real bummer when you're trying to track down
what commit introduced the breakage.
Does that explanation make it clearer why I think that build failures
are more serious than testsuite regressions?
Cheers,
Carlos.