Re: [Issue 806] Changed - De-kluge libsvn_ra_dav's fix for issue #797, use clean fix

Philip Martin <philip@codematters.co.uk> writes:
> Sounds OK. It sounds like you are storing them in memory, with the
> access baton pointing to them.

Right, that's what I was thinking.

> An alternative is to hold an open
> apr_file_t in the access baton, and write to that. Then
> svn_wc__run_log uses the apr_file_t, either as part of the log, or as
> the whole log. In fact, the entire log file could be accumulated this
> way.

Hmm, yes. It's a bigger rewrite of the way log appending works, since
currently the code accumulates logs in a stringbuf and then writes it
out all at once at the end.

But then again, the log file management is a bit messy right now. If
function A is accumulating a log, and calls function B which
accumulates and then runs a log, then returns to A, which writes out
*its* log and runs it, then B's log commands get run before A's, even
though one almost certainly wanted log commands to be run in the order
in which they were accumulated.

And look at libsvn_wc/update_editor.c/svn_wc_install_file()... It
seems to assume that it has the run of the log file, and can run it
when done. Maybe that's okay, I'm not sure.

Yes, there's a bit of mess here, and I'm not sure how much of it we
want to clean up as part of issue #806. Will sleep on it.

> This is a bit like another idea which is to have the access baton be
> aware of when a log file is running. Certain operations are allowed
> when running a log file that are prohibited at other times.

Care to elaborate? :-)

-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 26 00:49:48 2002