Urias McCullough wrote:
The build process includes testing that the result works as expected, and that pathological code isn't included. I hear you, there are two ways of looking at it, but the mechanical act of getting something to compile is kiddyland.On Tue, May 19, 2009 at 6:10 PM, Rob Judd <haiqu@xxxxxxxxxx> wrote:I just reloaded r30782 (which is the last RELIABLE build) and the problem disappeared.You have a different definition of "broke the build" than I'm familiar with... this usually represents a situation where the act of compilation fails.
No, I obviously haven't tested every build between r30782 and r30805. I did, however, report that the last build I attempted, at r30890, failed to install from CD. Since then I have changed to building on Haiku (vs Dano previously) on a completely different machine, so I have internal confirmation / double-checking.So, by indicating that 30782 was the last working build, you're saying that 30783 was the busted revision? This was a change to the nvidia driver... I don't have any nvidia-based haiku test boxes these days, so maybe someone else can confirm this?
I'll do some testing and see what comes out of that, but it sure is eating into my productive programming time. Right now I'm spending 50% of my available time keeping the O/S running and/or rebooting and cleaning up damaged files and/or reinstalling to new partitions when it all becomes unusable. It's fairly unstable.
Rob