For those interested in swapping in some other database backend
instead of Berkeley, I just spent the last 11 hours working towards
making that possible.
I estimate that I'm about 75-80% finished with a coding task that
abstracts away all the use of the skel_t structures in libsvn_fs to a
single source file of conversion/validation routines between real FS
structures like "svn_fs_representation_t" and the skel stuff, and then
the few source file responsible for writing to BDB. When my task is
finished, all the FS core code that doesn't directly write to BDB will
be talking in terms of regular C structures, not the skeleton
objects. That means that new databases that have real multi-column
table support write out multiple columns without parsing and unparsing
"list-y marshaled strings".
But for now, I must sleep. Must sleep.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu May 2 15:05:40 2002