Changes to ra_svn layer

Yesterday and today I did a bunch of ra_svn hacking. The upshot is:

  * Many incompatible protocol changes were made. Hopefully it's
    better to make them all at once instead of strung out across a
    long time.

  * All errors should be reported to the client now. (However, many
    of them will be reported later than they actually happen, due to
    pipelining.)

  * There should be many fewer round-trips. Report operations have no
    round trips; edits should have at most two round trips per changed
    file, plus a couple for each directory. (Because of the
    double-flushing conundrum, we don't buffer these pipelined
    operations very well. But the round-trip delays are still
    eliminated.)

I used the test suite, of course, but it's still possible that I broke
stuff. Let me know if I did, and I'll try to fix it promptly. (But
make sure you're running the same rev on the client and server, of
course.)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon May 26 22:40:49 2003