This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: More test results...


On March 8, 2002 01:15 pm, Benjamin Kosnik wrote:
> > I applied your patch out of the box (what the heck) and
> > bootstrapped, but I'm getting testsuite failures on
> > 18_support/numeric_limits.cc. For example,
>
> Yeah. I hadn't really run 'make check' when I posted that last
> night...I fixed this with the check in.

Ill do another bootstrap as soon as the load level on the CVS server 
goes down enough to check out the latest.

> Incidentally, how are you testing KDE?
>
> I see the KDE sources here:
> http://download.au.kde.org/pub/kde/unstable/kde-3.0-rc2/

I'd be happy to get them to compile at this point.  A successful 
compile takes many hours, assuming my hard drive doesn't max out part 
way through.  At the moment compiles fail on that vector<bool> problem.

I've been playing with the beta2 sources since that's what I've got.

> And the Qt sources here:
> ftp://ftp.trolltech.com/qt/

I have, through my employer, a commercial license for Qt so I have a 
later snapshot than the free version.  I've been testing version 3.0.2 
bacuse that's what I need.

> Are you just compiling them? Hmm.

I use Qt in day-to-day software development, so I've been giving it a 
fairly thorough testing from that perspective.

> The release critieria just has Qt on it. Due to the problems that
> gcc-3.0.x had/has with KDE, it would make sense to add those packages
> as well, I'd think. However, it might also be prudent to specify
> acceptance criteria....at least what "testing" and "pass" mean. Any
> thoughts?

At the moment I keep a 2.95 toolchain around for KDE stuff and Linux 
kernel rebuilds.  It would be nice if I could get rid of it.

KDE is a big effort to build, but it does come with its own testsuite.  
I'm not sure how expansive its test coverage is, but KDE itself 
certainly covers an awful lot of territory, and requires that Qt be 
buildable and working.

Since Qt is a commercial product it doesn't ship with a testsuite.  
Perhaps it's sufficient that KDE (rather than Qt) be made the release 
criterion, and that acceptance criteria include that it builds and that 
it passes its own testsuite with no regressions from an established 
benchmark such as GCC 2.95.  

That's no worse that the "builds and basic functionality" guideline.

-- 
Stephen M. Webb


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]