On 05/01/2012 03:00 PM, Danek Duvall wrote:
...
Some of the parfait errors in vim took me hours to track down, and I
want to make sure that that kind of analysis doesn't get lost, and
that the bugs that don't require such analysis get annotated with the
simple analysis that demonstrates how simple the bug is.
Fair enough. I've added a more detailed Evaluation to the Bugster CR.
I've also done the same for my other recent Parfait bug fixes:
7164358 Parfait problem with hexedit
7164360 Parfait problem with links
7164371 Parfait problems with snort
7164372 Parfait problems with gocr
I can agree with the fix by looking at it, but simply running the
program before and after the fix doesn't demonstrate that anything's
changed, never mind that a bug has been fixed. Part of your testing,
of course, will be to show that parfait doesn't complain after the
fix, but the other part is to demonstrate that the bug parfait found
(file descriptor not being closed) is no longer present -- truss
output showing that the close happens before main() returns, maybe?
Understood. I could have checked the return value of the newly added
fclose() call and made sure that is was good, but for this simple case,
the 'testing' of the fix here really was to eyeball the changes and see
that they were correct.
It's unfortunate that this (and the other four components above that
have been deparfaited) don't have any test suites. In some of the other
cases, it would take an inordinate amount of effort to cons up a test
case specifically to test each change.
But once we have agreed on what the changes should be, and they've
been integrated, then an attempt will be made to push them upstream.
Hopefully acceptance by the authors of the software will confirm we've
made the right changes.
I know this is a super-simple thing, and I don't want to overburden
you with a lot of paperwork for such a small problem, but I felt that
it would be easy to demonstrate on a simple case what ought to be done
for these bugs when they got more complicated.
For these relatively simple ones, the paperwork overhead is minimal and
I agree that it helps others who might have to make similar parfait fixes.
For some of the others components that have generated numerous Parfait
complaints, then it's going to get a little more burdensome.
Anyway, thanks for starting to tackle these guys. Danek
Sure. Can I take it that adding the Evaluation to the aalib bug report and
your previous "The fix looks fine" comment, that this one is now good to go?
Thanks.
_______________________________________________
userland-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/userland-discuss
