RE: [Refdb-users] Refdbib segmentation fault in 0.8.6

[Markus H. <hoe...@co...>]

Hi Nicolas,

Nicolas Brouard writes:
 > char verbose[PREFS_VUFS_LEN] = "f"; /* should be true (t) or false (f) */
 > 
Ok, I missed that one too. I've fixed all this stuff a while ago in the
libdbi-pre1 CVS branch that I mainly develop currently, but I hope
that I'll get it done for the 0.8.x series as well.

 > Refdbib has always worked with a call read_prefs(,,1)"
 > but you setup it with (,,0). Now the latter works if you supply arguments
 > (-S foo -t bibtex etc.) but if you only
 > type ./refdbib without arguments there is still a
 > segmentation fault. It is probably while reading the refdbibrc file. But
 > what is the difference between 0 and 1 and way some clients use 1 and
 > refdbib should use 0. It is not clear for me.

Only bib2ris uses read_prefs(,,1). This is a special case as bib2ris
uses a linked list to read a theoretically unlimited number of BibTeX
field to RIS field mappings from a configuration file. All other
bib2ris configuration data as well as the configuration data of all
other RefDB applications end up in arrays of fixed size as all
variable names to look for are known at compile time. Using
read_prefs(,,1) with refdbib in your experiments just so happended to
not crash. It is not supposed to work, as you can see if you take a
look at the implementation of read_prefs().

If running ./refdbib without arguments causes a segfault it has to be
fixed. This doesn't mean that calling refdbib without arguments would
make any sense, but it is still a bug. Unfortunately it doesn't crash
on FreeBSD, so I'll have to wait until I can access my Linux box again
to verify that this problem is fixed.

regards,
Markus

-- 
Markus Hoenicka <hoe...@co...>
http://ourworld.compuserve.com/homepages/hoenicka_markus/