Darren J Moffat wrote:
Danek Duvall wrote:
On Thu, May 14, 2009 at 10:27:55AM -0600, Mark J. Nelson wrote:
I'm kind of lukewarm on this idea in general, but don't really object.
I'll echo that, though adding that eliminating noise changes should
probably only eliminate noise that's gone away -- new noise should
probably
always be seen.
I'll take a much stronger stance than that and say that new noise MUST
be seen.
Part of the point here is that "new" noise, following a failed build, is
not really new. It's the normal build noise, which *should* have been
generated by the previous build, but wasn't.
This whole change though makes me very nervous as I suspect it is all
driven by a misguided desire for empty sections in the nightly mail. It
should be obvious reading the nightly mail when the build was clean.
Right. Hence my being lukewarm--as an advocate, I would not accept
build logs with use of this flag. Which is why I suggested not having
empty output, but instead explicitly saying "the -s flag was used." But
an empty noise section is already not indicative of a clean build--it
simply means that, between the previous build and the current build, no
new noise was introduced. :(
We already have some issues with the way we process noise in the build.
It's really easy to miss something the first time it's introduced, and
then you'll never see it again, or it will be utterly lost in a flood of
false information following a failed build. This fix seems to not
address any of those real issues.
If the behaviour this rfe is suggesting is acceptable why isn't it the
default case, or why isn't it the only case. IMO nightly is already too
configurable.
--
Darren J Moffat
_______________________________________________
tools-discuss mailing list
[email protected]
