general pool problem

Folks (Greg especially, since he has lots of APR experience):

I'm noticing a general problem, annoying though not fatal. It seems
that any function returning (svn_error_t *) must take a pool argument,
because the svn_error constructor needs a pool to allocate from.

This means that, like, virtually everything in Subversion has to take
a pool argument, even the delta-walker callback functions, for
example. It's not a huge pain, but it means adding an extra argument
to many functions.

Is this just the way it has to be? Is this the price of the
svn_error_t system? If so, do you think we would be better off using
statically sized errors from the user-reserved range of apr_status_t
values?

I like the wealth of information preserved by svn_error chains, and
would hate to give it up, but it is costing us in complexity somewhat.

-Karl
Received on Sat Oct 21 14:36:06 2006