On Tuesday 15 February 2011 02:15:18 Jeff Johnson wrote: > > (No rpm 5.3.8 working on my platform leads me temporary back to the > > beecrypt 4.1.2 problem). > > You can build --with-beecrypt=internal, that's an up-to-date 4.1.2 image > distributed with RPM. The routines will end up in -lrpmmisc.
I thought about that but the rpm 5.2.1 tarball I have doesn't provide beecrypt sources internally and it can only be configured to be external. > Berkeley DB (and PCRE) can also be put back internally, any version. Just > untar the sleepycat tarball and rename to "db", then --with-db=internal. Ok so I suppose it's the same for beecrypt. > Hmmm yes this is library loading. I have a dim memory of an asm > section/segment bug fixed in beecrypt like back in 2009 when 4.1.2 > was released. It also caused some issues with SELinux textrel policy. > > Again these are rather ancient issues, and not on arm. Just fyi ... Yes I think the reason is older kernels, not arm (I've seen this problem only with 2.6.28 and 2.6.29, but all works with 2.6.33). > I'm not as good at reading kernel panic's as I once was, but that "PREEMPT" > sure looks fishy ... and I'm not sure how benchbc gets clobbered by > beecrypt, its not like beecrypt is doing DMA or reading speshul kernel > drivers or anything other than just using optimized asm. > > You might try taking out the optimization when building older beecrypt, > there's likely some sneaky ASM instructions used even on ARM. I'm building with -O2, ok I can try without it, if that's what you mean. Thanks. Silvan -- mambaSoft di Calarco Silvan Web: http://www.mambasoft.it mambaSoft Store @ http://www.mambastore.it openmamba GNU/Linux development @ http://www.openmamba.org ______________________________________________________________________ RPM Package Manager http://rpm5.org Developer Communication List rpm-devel@rpm5.org
