On Thu, 26 Nov 2009, koettermarkus_at_gmx.de wrote:

> Yes, it even works.
> I verified BLKSIZE works too, added a new option to the curl cli.
> I tested the multi-interface as well.

Cool!

> * manpage docs for the --tftp-blksize option

I can volunter for that part.

> * recycling connections, I turned it off

Yeah, we should have them off. UDP transfers have basically no reasons for
getting reused so we are probably better off not even attempting.

> * connect() after we have received the first packet from the server, avoids
> problems as the kernel will take care we only get packets from the remote
> host.

Sounds like a good idea!

> * maybe indenting, would be handy if you could create a uncrustify.conf for
> curl.

Ah. I never even seen or heard of uncrustify before (I'm a GNU indent user
since ages though) but it looks like a capable tool so I should give it a go!

Personally I indent all code with emacs... :-)

> If it was svn, I'd commit it myself, as atomic commits allow easy reverts.

Yeah, but A) it's not _that_ hard to revert a CVS commit and B) we really
shouldn't revert it anyway but polish it up until it looks nice and shiny! And
the (B) part is easier to work interatively on after the initial commit has
been done.

I think you can commit it.

> It may break things, it may not compile on some plattforms, but it is more
> than a good start.

It really shouldn't break very much. It doesn't alter much if any platform-
specific stuff. And TFTP is already somewhat broken so I think a short period
of breakage until we fix things wouldn't be terribly bad anyway.

> diff is attached.

Unfortunately the diff wasn't -u!

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html