You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Robert Pluim <rp...@bigfoot.com> on 2003/04/07 17:16:03 UTC
svn diff segfault (issue 1093 again?)
Move a file:
rpluim: ~/wc-test/local-wc1/foo
$ svn mv 1000 1001-moved
A 1001-moved
D 1000
Add some stuff to it, run diff (BTW, since this now gives me only
local changes, perhaps issue 1061 is no longer relevant) :
$ svn diff 1001-moved
Index: 1001-moved
===================================================================
--- 1001-moved (revision 28)
+++ 1001-moved (working copy)
@@ -43,3 +43,4 @@
# compression = yes
### See http://subversion.tigris.org/issues/show_bug.cgi?id=668
### for what else will soon be customized in this file.
+more stuff
Everything seems ok, except:
rpluim: ~/wc-test/local-wc1/foo
$ svn diff -r 27 1001-moved
Segmentation fault (core dumped)
rpluim: ~/wc-test/local-wc1/foo
This is coming from subversion/libsvn_client/diff.c:1227, where we do:
SVN_ERR (convert_to_url (&URL1, path1, pool));
/* Trickiness: possibly split up path2 into anchor/target. If
we do so, then we must split URL1 as well. We shouldn't go
assuming that URL1 is equal to path2's URL, as we used to. */
SVN_ERR (svn_wc_get_actual_target (path2, &anchor, &target, pool));
if (target)
{
svn_path_split (URL1, &url_anchor, &url_target, pool);
And surprise, surprise, URL1 is NULL, since path1 was a moved file.
I'm not sure of the best way to handle this. convert_to_url can't
give me the actual previous URL, because that's rev-dependent. Giving
back the URL as it was just prior to the move would work in 90% of the
cases, but would fail dismally if the file has been moved more than
once. Issue 1093 seems to classify this as an 'annoyance', so maybe I
should close my eyes and ignore it :)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn diff segfault (issue 1093 again?)
Posted by Karl Fogel <kf...@newton.ch.collab.net>.
Robert Pluim <rp...@bigfoot.com> writes:
> http://subversion.tigris.org/issues/show_bug.cgi?id=1229 done sir.
> There doesn't seem to be a pre-1.0 milestone, should I set it to
> "Beta"?
Yup, that's what I meant, sorry.
Thanks!
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn diff segfault (issue 1093 again?)
Posted by Robert Pluim <rp...@bigfoot.com>.
Karl Fogel writes:
> No, a segfault is never ignorable :-). Can you file an issue for
> this? It should be pre-1.0, and separate from 1093 (though it may
> want to reference 1093)...
http://subversion.tigris.org/issues/show_bug.cgi?id=1229 done sir.
There doesn't seem to be a pre-1.0 milestone, should I set it to
"Beta"?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn diff segfault (issue 1093 again?)
Posted by Karl Fogel <kf...@newton.ch.collab.net>.
Robert Pluim <rp...@bigfoot.com> writes:
> Everything seems ok, except:
>
> rpluim: ~/wc-test/local-wc1/foo
> $ svn diff -r 27 1001-moved
> Segmentation fault (core dumped)
> rpluim: ~/wc-test/local-wc1/foo
>
> This is coming from subversion/libsvn_client/diff.c:1227, where we do:
>
> SVN_ERR (convert_to_url (&URL1, path1, pool));
>
> /* Trickiness: possibly split up path2 into anchor/target. If
> we do so, then we must split URL1 as well. We shouldn't go
> assuming that URL1 is equal to path2's URL, as we used to. */
> SVN_ERR (svn_wc_get_actual_target (path2, &anchor, &target, pool));
> if (target)
> {
> svn_path_split (URL1, &url_anchor, &url_target, pool);
>
> And surprise, surprise, URL1 is NULL, since path1 was a moved file.
>
> I'm not sure of the best way to handle this. convert_to_url can't
> give me the actual previous URL, because that's rev-dependent. Giving
> back the URL as it was just prior to the move would work in 90% of the
> cases, but would fail dismally if the file has been moved more than
> once. Issue 1093 seems to classify this as an 'annoyance', so maybe I
> should close my eyes and ignore it :)
No, a segfault is never ignorable :-). Can you file an issue for
this? It should be pre-1.0, and separate from 1093 (though it may
want to reference 1093)...
-K
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org