Thanks. I've debugged a little, and it seems that the primary reason is that a bogus value is read for n_lines in read_documents (file sfm_read.c).
Once this has failed, the code paints itself into a corner by checking whether the file was already open in sfm_read_dictionary; this test succeeds, but the content of the dict structure remains bogus... I can't reproduce your segfault, but as I said, this phenomenon is probably secondary to the real problem anyway.
The value that is getting read for n_lines is 134217728 aka 0x8000000 which looks like it could be intended as a special value?
I've got no idea whether it is a special value, I'm pretty sure I didn't select anything special myself when saving that file, a couple of years ago, because I also always kept the back then recommended "more portable" .sys files. All I know is that I never had problems before R 1.7.1 and that the .sav files reads correctly in SPSS 10 on Windows (and 9), the last SPSS for Mac OS 9 (v10?) and the new SPSS v11 for Mac OS X.
- Hedderik.
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help
