Re: segfault checking out over ra_dav

Brian Denny <brian@briandenny.net> writes:

> gdb backtrace follows. further investigation reveals that CHECKSUM
> (dereferenced at fetch.c:739) is NULL.
>
> any ideas? can anyone reproduce this?
> i can't give an ethereal trace at the moment, but maybe tonight...

You can use the client to get a trace. Set "neon-debug-mask = 130" in
~/.subversion/servers you will get a neon debug trace.

> if somebody wants to try checking out from my repository with your local
> client, try:
> svn co http://frondle.net:5050/svn/repos/test

I get a SEGV as well.

It appears that your server isn't sending md5-checksums. If I use my
local server then for each file the neon trace contains a line like

<lp2:md5-checksum>d3b07384d113edec49eaa6238ad5ff00</lp2:md5-checksum>

but the trace from your server doesn't contain any of these. I don't
know if this is supposed to happen, I guess it can if it's an old
repository without checksums.

The following patch appears to make the client work, but I know next
to nothing about ra_dav.

Index: subversion/libsvn_ra_dav/fetch.c
===================================================================
--- subversion/libsvn_ra_dav/fetch.c (revision 5446)
+++ subversion/libsvn_ra_dav/fetch.c (working copy)
@@ -736,7 +736,9 @@
      simple_fetch_file() params related to fetching version URLs (for
      fetching deltas) */
   err = simple_fetch_file(sess, bc_url, NULL, TRUE, compression, file_baton,
- NULL, checksum->data, editor, NULL, NULL, pool);
+ NULL,
+ checksum ? checksum->data : NULL,
+ editor, NULL, NULL, pool);
   if (err)
     {
       /* ### do we really need to bother with closing the file_baton? */

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org