Bug 165465 - svn can segv on export of working copy
Summary: svn can segv on export of working copy
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: subversion
Version: 4
Hardware: All
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Joe Orton
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2005-08-09 16:33 UTC by Nigel Metheringham
Modified: 2007-11-30 22:11 UTC (History)
0 users

Fixed In Version: 1.2.3-2.1
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2005-09-07 08:01:28 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)

Description Nigel Metheringham 2005-08-09 16:33:34 UTC
Description of problem:
  $ svn export . /tmp/eview
  Segmentation fault

Version: subversion-1.2.1-2.1


How reproducible:
  On this working copy (which has been in use for a few months), every time.
  If I check out the data again then it works OK.
  Differences between working copy and new checkout are some minor meta-data
  changes (ie prop-time & text-time changes).


This is likely to be a bit of a heisenbug.
I have the affected directory tree around in case there are some diagnostics you
wish to perform on it.
strace is giveing no clues I can see - last few lines are:-
 write(4, "#!/usr/bin/perl\n#\n#\t$Id: upload_"..., 1594) = 1594
 close(4)                                = 0
 rename("/tmp/eview/evlogana/bin/upload_queue.pl.tmp",
"/tmp/eview/evlogana/bin/upload_queue.pl") = 0
 stat64("/tmp/eview/evlogana/bin/upload_queue.pl", {st_mode=S_IFREG|0664,
st_size=1594, ...}) = 0
 chmod("/tmp/eview/evlogana/bin/upload_queue.pl", 0775) = 0 
 stat64("/tmp/eview/evlogana/bin/upload_queue.pl", {st_mode=S_IFREG|0775, 
st_size=1594, ...}) = 0
 utimes("/tmp/eview/evlogana/bin/upload_queue.pl", {1123604392, 0}) = 0
 --- SIGSEGV (Segmentation fault) @ 0 (0) ---
 +++ killed by SIGSEGV +++

Comment 1 Joe Orton 2005-08-09 16:41:22 UTC
Can you get a backtrace from the crash?  (install subversion-debuginfo, gdb svn,
"run export . blah", "backtrace full"?)

Comment 2 Joe Orton 2005-08-09 16:49:09 UTC
Never mind, this is fixed upstream for 1.2.2.

------------------------------------------------------------------------
r15579 | dlr | 2005-08-03 23:20:13 +0100 (Wed, 03 Aug 2005) | 6 lines

Merge r15516 from trunk for the 1.2.2 release, fixing a 'svn export
wc_path' client SEGV.

Votes:
+1: philip, sussman, dlr


Comment 3 Joe Orton 2005-08-09 16:53:44 UTC
Can you confirm that there is a file in the "deleted" state (i.e. an uncommitted
'svn rm' or 'svn mv') in the working copy which triggers the crash?

Comment 4 Nigel Metheringham 2005-08-10 08:08:43 UTC
There are no uncommitted changes within the tree.  [This was showing up as part
of my checkout and build an rpm script, so does a ci, tags the repo, exports,
tars & rpmbuilds.  The working copy will be in a clean state].

Just to be really annoying I cannot reproduce this problem today - yesterday of
course I couldn't not reproduce it.

Suggest this be pushed into a quiescent state and I'll re-open if I manage to
reproduce it.  Unfortunately it came up just around the time I needed to leave
in the evening so I didn't have time to collect backtrace info.  I have seen a
similar heisenbug style segv with export in a different svn build.

Comment 5 Joe Orton 2005-08-30 10:54:38 UTC
The 1.2.3 updates in testing have the "svn export" fixes mentioned above; you
could try those.

  # yum --enablerepo=updates-testing update subversion

Comment 6 Fedora Update System 2005-09-06 05:38:04 UTC
From User-Agent: XML-RPC

subversion-1.2.3-2.1 has been pushed for FC4, which should resolve this issue.

If these issues are still present in this version, then please re-open this bug.


Comment 7 Nigel Metheringham 2005-09-06 10:22:40 UTC
Confirm subversion-1.2.3-2.1 appears to fix the problem for me.

Comment 8 Joe Orton 2005-09-07 08:01:28 UTC
Thanks.


Note You need to log in before you can comment on or make changes to this bug.