On 21 August 2007 at 11:34, Ben Goodrich wrote: | -----BEGIN PGP SIGNED MESSAGE----- | Hash: SHA1 | | Dirk Eddelbuettel wrote: | > I still have the feeling that I am not quite sure why and where this would be | > a bug in R. As Doug demonstrated, it works merrily on amd64 with the same | > toolchain. | > | > So the ball is back in your court as we're not quite sure where this is a bug | > in R. Can you strip down the program you are trying to run? What happens, | > say, if you use valgrind on 'r' from the littler package (which initializes R | > to the minimal extent) ? | | I am not sure where the bug is either :) . For littler, I get a similar | error as with regular R 2.5.1 . I just compiled R-devel from the vanilla | source to compare with Doug (just ./configure CC=gcc-4.2 | - --with-valgrind-instrumentation=2; make). I got a crash, but with
Well I guess Doug may not have used '--with-valgrind-instrumentation=2'; our packages don't either. Apart from that, I am also lost. Maybe we need to talk to the valgrind or gcc maintainers? | R-devel, it does not have the analogue to the line | | - --27909-- Reading syms from /usr/bin/r (0x400000) | - --27909-- object doesn't have a symbol table | - --27909-- Reading syms from /lib/ld-2.6.1.so (0x4000000) | | that littler and R 2.5.1 produce. With R-devel, it appears not to have a | problem until a couple of lines later: ld-2.6.1 reminds me -- this may be a library issue. I think we just jumped from libc6 2.4 to 2.6, and not all compilers may be aligned. Just a random thought though. Dirk (rest trimmed) -- Three out of two people have difficulties with fractions. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
