On Wednesday 12 May 2010 09:05:17 Denys Vlasenko wrote: > On Mon, May 10, 2010 at 2:28 AM, Denys Vlasenko > > <[email protected]> wrote: > > Also, there is scripts/randomtest[.loop] scripts which I use for > > randomconfig build tests. They are not usable as-is, but can be tweaked. > > I will look into it. I think I will tweak it so that it takes > > source directory as $1 instead of hard-coding it as "tree", > > and uses $LIBC and $CROSS_COMPILER_PREFIX from environment. > > > > I do use randomconfig build test fairly regularly, so it should not > > error out - that would be a regression. > > > > Unfortunately, I did not run the testsuite against the resulting build, > > thus there are sure tons of cases when it will fail. > > I fixed all false positives and two real bugs which I found. > Now, this runs for 300 iterations without exiting on build > or testsuite error: > > busybox.git/scripts/randomtest.loop busybox.git > > (assuming busybox.git is a pristine git tree). > > You wouldn't want to use randomtest.loop directly in your nightly > test builds, I suppose. Just take a look at its code, and write > something along those lines which suits your environment.
For the cross-architecture stuff, I suspect defconfig or allyesconfig are the best things to test. What randconfig is really testing is that all the config options line up properly, which isn't really a platform-specific thing. (It's hard to get that wrong just on PowerPC without getting it wrong on x86 too.) I'll try to find time to set something up this weekend. Thanks, Rob -- Latency is more important than throughput. It's that simple. - Linus Torvalds _______________________________________________ busybox mailing list [email protected] http://lists.busybox.net/mailman/listinfo/busybox
