On Sun, 5 Jan 2014, Daniel Stenberg wrote:

> I believe the reason is that LIST is done after TYPE A is sent by
> libcurl (ascii mode), and then it converts CRLF to LF before
> storing the data and then the comparison has to be done
> without the CRs. On windows, converting CRLF to plain LF is
> probably to be considered the wrong thing...

The conversion on Windows doesn't seem right to me based on my knowledge of
the email protocols although I guess there could be legitimate reason within
FTP.

I have often thought whether we should introduce a EOL setting in the <data>
element which would then be used in the comparison rather than the line
endings that are at the end of each line. This would be needed if we were to
fix the conversion on Windows so that the tests succeed on each platform. It
could then be set to CRLF, LF, CR and native. For example:

<data eol='native'>
some data content
</data>

> My primary reason for the test change is however to be able
> to write tests for the FTP list parser!

I agree the change makes sense - it brings the FTP LIST command in line with
how the email tests perform similar actions.

The main reason for me bringing this up was to ask if we can fix those other
tests, that I mentioned in my earlier email, in the same manner as Test100
by simply changing the <datacheck> element to a <data> element?

Kind Regards

Steve

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2014-01-05