www.libssh2.org | Daily snapshots | Mailing list archive | Docs | Examples | github

Archive Index This month's Index

Subject: Re: libssh2 MorphOS patch

Re: libssh2 MorphOS patch

From: Leif Salomonsson <dev_at_blubbedev.net>
Date: 02 Aug 2012 10:00:02 +0100

Hello Daniel,

On 2012-08-01, you wrote:
> On Wed, 1 Aug 2012, Leif Salomonsson wrote:

> Thanks for your work!

>> Because of collissions with bsdsocket send() and recv() functions, some
>> changes where made in libssh2_priv.h and session.c, where these names
>> where used for callbacks variables in LIBSSH2_SESSION.

> I don't get why those "collissions" cause any problems. Do you have them
> as macros/defines or something? Why do they collide? Every single
> operating system we ever build libssh2 on have send, recv etc...

Yes these functions are defines on MorphOS.

> Your changes indicate there is something seriously wrong with your
> headers.
> Further, your indents and white space changes were off. We use
> indent-level as 4 spaces and no tabs anywhere.

Ok, noted.

>> Similar issue in agent.c, where ->connect() is changed to ->_connect_().

> Perhaps, but you also changed the *actual* connect() call to _connect_()
> which surely doesn't work?

No, just changed ->connect().

>> Some blocking mode set/get changes/fixes in session.c.

> I don't like how everything MorphOS there is within HAVE_IOCTLSOCKET_CASE
> when they obviously are far from related to that function alone. For
> example:
> +#ifdef HAVE_IOCTLSOCKET_CASE
> +#include <proto/socket.h>
> +#endif

> .... should rather be:

> #ifdef HAVE_PROTO_SOCKET_H
> #include <proto/socket.h>
> #endif

> And this:

> +#if defined(HAVE_IOCTLSOCKET_CASE) && (GETBLOCK == 0)
> + /* Amiga/MorphOS */
> + unsigned int option_value;
> + unsigned int option_len = sizeof(option_value);
> +
> + if (getsockopt(sockfd, SOL_SOCKET, SO_ERROR, (void *) &option_value,

> &option_len)) {
> + /* Assume blocking on error */
> + return 1;
> + }

> .... I think should rather use #ifdef _MORPHOS or something as it seems
> more related to the OS rather than the actual way ioctlsocket() is used.

My reasoning is that this way it works also for other Amiga-style
OS'es. As the (cased) IoCtlSocket() function is afaik unique
to the bsdsocket implementation shared between theses OS'es.

> BTW, MorphOS is not present here and I'd appreciate if you'd add it:
>
https://sourceforge.net/apps/mediawiki/predef/index.php?title=Operating_Systems

Ok.

Regards,
Leif

_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2012-08-02

the libssh2 team