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

Re: [neon] [azverkan@yahoo.com: Joe Orton please read: Overrrun a neon buffer size during svn merge URL URL]

From: Joe Orton <joe_at_manyfish.co.uk>
Date: 2002-12-20 23:50:11 CET

On Fri, Dec 20, 2002 at 02:11:15PM -0800, Matt Kraai wrote:
> On Fri, Dec 20, 2002 at 10:04:33PM +0000, Joe Orton wrote:
> > On Fri, Dec 20, 2002 at 06:19:01AM -0800, Kevin Pilch-Bisson wrote:
> > > While running svn merge URL URL, we overrun a 4K buffer size max in neon
> > > and this caused subvesion to exit with an error before finishing its
> > > work. The buffer in question is in new_socket.c line 164 "#define
> > > RDBUFSIZ 4096". After upping this value by multiplying it by 32 the
> > > operation finished successfully.
> >
> > Can you explain why this is a problem, what causes this failure (with a
> > debug trace)? neon should work correctly (albeit not optimally) with any
> > size of read buffer.
>
> It would be harder to run into the "Line too long" error from
> ne_sock_readline if RDBUFSIZ was larger, but some specification may
> state that 4096 is sufficient.

Is that the only reason why you want the RDBUFSIZ to be increased? What
is the exact error being returned?

ne_sock_readline is used for reading the status-line and chunk size
lines, for both of which it is quite reasonable to pick a maximum
acceptable length of 4096 chars.

The other time readline is used is for reading response header fields:
I'd question whether it's reasonable for a response header field to be
longer than 4096 chars too. Is mod_dav_svn really sending these in some
case? Please send network/debug traces!

Regards,

joe

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Dec 20 23:50:44 2002

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.