On Tue, 19 Feb 2002, Blake Girardot wrote:

> I built cURL on a YellowDogLinux box:
>
> curl --version: curl 7.9.3 (powerpc-unknown-linux-gnu) libcurl 7.9.3
> (OpenSSL 0.9.6)
> uname -a: Linux 2.4.10-12a #1 Tue Oct 9 04:29:39 EDT 2001 ppc unknown
> bash --version: GNU bash, version 2.04.11(1)-release
> (powerpc-yellowdog-linux-gnu)
>
> YellowDogLinux is basically a RH 7.1 if I understand things correctly.

First, for the record, I am not aware of any FTP bugs in recent curl versions
that would make this happen, unregarding of operating system. I've not seen
this before.

*doh* It strikes me here and now. I think this might be a 7.9.3 bug I
corrected in 7.9.4 (dated January 29, 2002: "The upload byte counter
bugged").

Could you upgrade and retry? I'd recommend the 7.9.5-pre1 version, I think it
is better than the 7.9.4...

> curl -u $FTPUSER:$FTPPASS -T$BACKUPDIR/$COMPUTER-DAILY-$YEAR$NOM$DOM.tgz \
> ftp://$FTPSERVER$FTPPATH$COMPUTER-DAILY-$YEAR$NOM$DOM.tgz
>
> Everything goes basically ok, but curl returns something like this every
> single time:
>
> curl: (18) Wrote only partial file (11783908 out of 11691024 bytes)

This is exactly 92884 bytes more. Is the reported numbers always the same? I
figure the exact sizes differ since you seem to upload a dynamicly created
file, but what about the size difference, how does that vary?

> You will notice it only wrote _more_ than the file size, and counted this
> as a partial transfer.

Yeah, the check in the code is for differences, it is just the error text
that should be put differently in this case. It is still considered an error
internally.

-- 
    Daniel Stenberg -- curl groks URLs -- http://curl.haxx.se/