On Wednesday 28 December 2011 23:26:04 Yang Tse wrote:
> ftpserver.pl was certainly not ready for the arbitrary splitting of
> application data among different TCP packets. I've just pushed some
> changes to ftpserver.pl that should make pingpong server it work with
> any sort of application data fragmentation it receives over TCP.
>
> Please verify if this commit alone improves/fixes
> https://bugzilla.redhat.com/760060

No, it does not. FTPS tests are still skipped in the same way as before.

> If e99128a5c9 does not improve the situation, logs from a manually run
> of "runtests.pl -k 100" and interrupting it should better pinpoint

What should be the connection between test100 and FTPS?

> where the problem comes from. I believe that making sockfilt.c break
> on new-line boundaries should not be required unless we have another
> problem somewhere else.

How else would we know we have read enough from the socket?

Please explain what the issue with my original patch actually is.

- it solves the problem that NSS introduced
- it does not break anything at first glance
- it does not change the code much

Kamil
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2011-12-29