Re: subversion 1.9.3-1 segfault

[Greg Chicares <gchicares@sbcglobal.net>]

On 2016-02-19 16:28, David Rothenberger wrote:
> Greg Chicares wrote:
[...]
>> /lmi/mirror/lmi[1]$svn update
>> Updating '.':
>> zsh: segmentation fault (core dumped)  svn update
[...non-helpful stackdump...]
>> I'll try running 'svn' under 'gdb'. Is there anything else I can do to help?
> 
> That's the only other thing I can think of. If you can catch the 
> segfault and get a backtrace, I may be able to track it down.

Thanks. The problem is no longer occurring. Let me state what I
know for the record in case it helps someone else someday.

This seems to be a heisenbug. When I ran it under gdb, repeatedly,
it refused to segfault. Outside gdb, with subversion-debuginfo
installed, I was unable to get a useful stackdump.

The problem has occurred only when gnu.org's svn server is slow:
on one isolated occasion a week or two ago, and throughout today.
And today there seems to have been a DDOS attack on gnu.org:
  https://savannah.gnu.org/support/?108973

While the problem lasted, sometimes I got one or both of these
messages:
  svn: E170013: Unable to connect to a repository
  svn: E175012: Connection timed out
and sometimes svn segfaulted with no message. I tried the same
operations on my debian-7 ("wheezy") system with its older
svn-1.6.17, where I got timeouts but no segfault.

My real problem was the gnu.org server. I can only conjecture
that attempting to work with a server under extreme load causes
svn-1.9.3-1 to follow some exceptional code path that attempts
to report a diagnostic and in doing so sometimes crashes.


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple