budd@arizona.UUCP (tim budd) (08/08/84)
I posted a note about this once before, but judging from the amount of mail i have been receiving the message has not gotten around much. In the 4.2 distribution tape, the gnomes at berkeley made several innocuous and two deadly changes between the time I sent them my final sources and they cut the tape. The deadly changes resulted from the fact that they use sccs, and I don't. Therefore two %D%D formats were converted by sccs into a date. I won't summarize the innocuous changes, but here is a diff script for the deadly ones. --tim budd, the university of arizona, arizona!budd. diff /usr/src/new/bib/src/invert.c /usr/src/local/bib/invert.c 97a95 > exit(0); 162c160 < sscanf(line,"%s%s83/09/23D", key, file, &start, &length); --- > sscanf(line,"%s%s%D%D", key, file, &start, &length); 165c163 < { sscanf(line,"%s%s83/09/23D", newkey, newfile, &start, &length); --- > { sscanf(line,"%s%s%D%D", newkey, newfile, &start, &length); another almost deadly change is to remove all the #%###!! sccs garbage from the front of files in the macs area. bib commands start with #, and everything after the first non-bib command is copied directly to output - I have no idea how berkeley got their system to work (perhaps they don't test the version on the distribution tape). ALSO - to answer other frequently asked questions - the file /usr/dict/papers/INDEX (or /usr/src/dict/papers/INDEX) is supposed to be a ``system wide'' database of commonly used references. It is searched automatically if a reference is not found in a local database. The database is created using invert, like any other database. The location the database can be changed by modifying a define in bib.h and recompiling. several people have complained about error messages resulting if this file is not found, however all this is described in the documentation. if you are just dragging the bib posted to the net along with a 4.2 upgrade, then you should be aware that 4.2 now allows long names, and thus the macros bibinc.shortnames, bibinc.fullnames and bibinc.publishers should be mv'ed from their 4.1 locations bibinc.shortna bibinc.fullna and bibinc.publishe . listrefs does not automatically list everything in an INDEX, indeed, listrefs never even looks at INDEX. The arguments given to listrefs must be a file containing the original references. See the man page for futher information. There have been some problems encountered in transferring to non 4.2 systems (system V, pe machines, etc). but nothing major enough to warrent reposting the sources.