[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: RA-DAV tests acting *very* strangely.

From: <cmpilato_at_collab.net>
Date: 2003-03-07 11:20:01 CET

cmpilato@collab.net writes:

> > Continuing to debug...
>
> Oh, geez. It seems that my revision 5202 commit broke this. Gonna
> spend a few moments trying to figure out why, but if I can't, I'll
> have to revert (and re-open that issue).

Hm. Well debugging led me to believe that 5202 didn't actually break
something, but perhaps shed light on a previously existing other
problem. I went over that change with a fine-toothed comb, and could
find absolutely no fault with it.

So I took my issue 1003 branch and ran basic_tests 1 over DAV. All
good. Then, I applied 5202's changes to that working copy. Now,
basic_tests 1 failed the same way my /trunk working copy was. After
some debugging, I realized I could make the crazy svn_io_dir_read()
behavior stop if I #if 0-d out this section of libsvn_client/commit.c:

  /* Get a readable stream of the file's contents. */
  contents = svn_stream_from_aprfile (f, pool);

#if 0
  /* Get an editor func that wants to consume the delta stream. */
  SVN_ERR (editor->apply_textdelta (file_baton, NULL, NULL, pool,
                                    &handler, &handler_baton));

  /* Send the file's contents to the delta-window handler. */
  SVN_ERR (svn_txdelta_send_stream (contents, handler, handler_baton, pool));
#endif

  /* Close the file. */
  apr_err = apr_file_close (f);

The test failed, of course, because no file contents were being
imported. But at least the weird stack-smasher-ish bug didn't occur.
So it seemed that one of those two function calls triggered the bug.

Unfortunately, I had no luck debugging what was between that
#if-#endif.

I tried to turn on pool debugging with "lifetime" enabled, but I think
there are bugs in the implementation of some of those pool debugging
params (Sander, feel free to ping me about this). By the time I'd
un-configured for pool debugging, I could no longer reproduce my
original bug. :-(

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Mar 7 11:21:58 2003

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.