Re: Problems with r15909 (ra_svn connect timeout)

On Thu, 2005-12-08 at 22:15 +0000, Julian Foad wrote:
> > In my opinion, yes: Apple's fixed the problem in their OS, so we shouldn't
> > be trying to hack around it.
>
> The criteria for back-porting are: it's an important bug fix, and it's safe
> (unlikely to cause problems). Does it meet those?

I think so, yes. A 30-second timeout in ra_svn is a significant
regression from 1.2.x, since operations can easily take longer than
that. And it's safe in that r15909 is a self-contained change made
since 1.2.x was released, and reverting that change should be as safe as
1.2.x was. Note that CHANGES needs to be amended to remove this line:

    * fixed: non-interruptible OS X socket connection with svn:// (r15909)

Malcolm, thanks for verifying that the OSX bug was real. I don't feel
it's really appropriate for applications to work around a bug of that
size, given how basic of an operation was broken and how many
applications perform blocking connects.

Karl, a fixed r15909 which unsets the timeout after the connect would be
a substantial improvement, but I still feel that simply reverting r15909
is superior. Thank you for reverting it.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Dec 8 23:38:31 2005