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

Re: Very slow svnserve responses on private repositories

From: Phil Bordelon <phil_at_lsu.edu>
Date: 2006-12-05 20:16:39 CET

Mark Phippard wrote:
> On 12/5/06, Phil Bordelon <phil@lsu.edu> wrote:
>>
>>
>> I'm using svn 1.3.2, both client and server, on Gentoo. neon is 0.26.1,
>> APR is 0.9.12.
>>
>> Please let me know if you need any more information to test this; it's
>> trivially reproducible on my machine, and I'd be willing to give a dev
>> an account on it to test on their own if that sort of thing would help.
>>
>> Thanks a bunch for such a great piece of software! I keep up with the
>> -dev list as much as I can, even though most of it goes over my head. ;)
>
>
> This should really have been sent to the users@ list.

I did, well over a month ago, and got no response:

    http://svn.haxx.se/users/archive-2006-11/0080.shtml

I read through the "rules" for the lists here and it says that you
should send to dev@ if you get no response and you think it's a bug ...
which I did. Apologies if that was the wrong way to go about it, but
if so, the guide should be updated:

    http://subversion.tigris.org/mailing-list-guidelines.html

That said, given the responses below, obviously I should have dug
through the user@ archives more before I posted. I apologise for
that.

> This sounds a lot like the fairly common svnserve problem where it is
> blocking in APR waiting for a random number to be generated from
> /dev/random. If that is the problem, then APR can be reconfigured and
> built
> to get the random number from /dev/urandom which will solve the problem. A
> lot of the people I have seen have this problem are using Gentoo. Here is
> an example:
>
> http://svn.haxx.se/users/archive-2006-11/0014.shtml
> http://svn.haxx.se/users/archive-2005-02/0828.shtml
>
> All that being said, the problem with /dev/random is a lack of entropy, and
> if the repository is on your desktop it seems unlikely that you would have
> that problem. It is worth a try though.

This does, indeed, fix the problem. I'm not using APR for anything that
needs deep cryptographic complexity, so I'm okay with using /dev/urandom
for my APR compile, but that seems an, ah, unobvious fix. In this
message:

    http://svn.haxx.se/users/archive-2005-02/0930.shtml

Karl suggested updating the documentation to cover this case. Consider
this a hearty vote in assent.

> Mark

Phil

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Dec 5 20:16:53 2006

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