Subject: Re: Allow authentication to be passed in memory - blast from the past

Re: Allow authentication to be passed in memory - blast from the past

From: David Calavera <david.calavera_at_gmail.com>
Date: Sun, 14 Sep 2014 16:20:16 -0700

Thanks for the comments Daniel.

The idea seems reasonable to me, but I would prefer if you'd post the
> patch(es) for review here on the mailing list rather than pointing out
> something (on github) so that we can comment and discuss them here.

I'm posting the code at the end of this email. I thought it would be easier
to read in a slightly richer view, since it's a slightly big patch, but I
totally understand it.

I would also wouldn't mind to hear someone else voice their support for
> this functionality just to be sure that we're working on something more
> people actually find sensible to add.

For what I've seen, someone needed this back in the day, that's where I got
the patch from
http://www.libssh2.org/mail/libssh2-devel-archive-2012-01/0015.shtml

I also saw an initial attempt to something similar in the trac:
http://trac.libssh2.org/ticket/248

I'd say that other people would find it useful, specially those that use
libssh2 as a dependency for their libraries. I'm using it with libgit2 and
the maintainers seemed interested when I talked to them.

I'd love to hear from more people too.

Personally I don't like the use of a macro that is the most widely used
> version of a function that then calls the actual *_ex() function - but I'm
> fully aware that we have a busload of functions done like that already...

It's funny because the original patch didn't have this macro. I added it
because I saw the rest ones :)

Does it also work when built to use gcrypt instead of OpenSSL?

I didn't try. Thanks for the reminder. I'll build it to use gcrypt and I'll
let you know if I find any issue.

This is the patch for everyone else to review:
Received on 2014-09-15