Peter Dimov said: (by the date of Mon, 6 Nov 2006 03:44:00 +0200)
Anthony Williams wrote:
"Peter Dimov" <pdimov@mmltd.net> writes:
It's not an expected failure if the optimization is not enabled.
I wouldn't know that if you hadn't just told me, and the summary report shows it as "Broken".
That's because the summary report has no way of knowing that the last known good release has been using a different configuration, as it only looks at the toolset name - AFAIK. It could be made smarter... or you could pick a different toolset name such as vc71-release so it can be marked up appropriately. Or is it already possible to mark failures based on debug/release? I don't know. In general I don't like marking tests that work as expected failures because when they fail for real, nobody notices.
Is is possible to mark expected failures as "with exactly this error message" ? If the message is different then it is no longer an expected failure but a real failure and people should notice. By that way regression tests will only allow this certain failure where the "linker is folding duplicate identical functions into one". -- Janek Kozicki |