On Fri, Apr 15, 2011 at 12:45 PM, brian m. carlson < [email protected]> wrote:
> On Tue, Apr 12, 2011 at 10:11:16PM -0400, A E [Gmail] wrote: > > Hi Brian, > > This is what I see [Notice the bold bit, which says, Signal 10, Bus > Error] > > *Program terminated with signal 10, Bus error.* > > Yes, this would be SIGBUS. It's an unaligned access, which means that > the software in question is buggy[0]. In C and C++, code which would > result in unaligned accesses is forbidden by the relevant language > standard. In contrast to the situation on some other processors, Linux > for SPARC does not have a way to automatically fix up unaligned > accesses, so the software has to be fixed. > > Just what I was afraid of, as I'm having a really hard time in having the developers to pay any attention to this as 'sparc' isn't a supported platform since they don't have any 'sparc' machines to work on. So I offered to donate a Sparc machine to the project with Debian 6.0.1 loaded on it, so they always have a sparc machine to test the releases on, or at least be able to test it if someone reports a problem. But that wasn't much to motivate them either. And unfortunately, I'm not a developer by any stretch of the imagination so I can't go into the code to figure out howto fix this. My only choice now is either they fix this OR I can figure out how to create 64-bit ssl libs that the make process finds compatible and moves the build along. I even sent emails to the openSSL list, but a week has gone by and no one has responded to it either. > > #0 aes_encrypt (plaintext=0xcdcd0, exp_key=0xcdcfc) at > > crypto/cipher/aes.c:1916 > > 1916 v128_xor_eq(plaintext, exp_key + 0); > > I expect that this uses some sort of vector operations but the data is > not properly aligned. Since all UltraSPARC machines support VIS 1 and > you're compiling for UltraSPARC (either explicitly or implicitly with > the Debian defaults), GCC may be using them even if you haven't > explicitly specified to do so. I'm not very familiar with the > intricacies of SPARC assembly, so I can't really tell you more. > > [0] Strictly, it could be instead that at that immediate moment you > pulled your hard drive (or some other essential system component, like > memory) out of the system, but I presume you would have mentioned that > if it were the case. > > Nope, nothing was installed or removed at that point. I have rebuilt this machine with a _fresh_ debian install from scratch about 3 times with a new git pull of the source code and everything and I get the same results every time :(
