mkoctfile urgently needs attention before Octave 3.6.0 release

[Lukas Reichlin]

I'm not in the position to block the release cycle. But I really think that 
Octave 3.6.0 must not be released unless regression #35114 [1] is fixed. During 
the discussion about my bug report [1], this regression turned out to be 
questionable feature! I do respect the developers' decision, but they must 
provide a convenient solution for the broken packages (at least 3 packages) on 
all platforms. This is not the case and I am not amused to say the very least.

After all, it's not one of these simple, long-standing bugs. It's a serious 
regression that renders all my work since 2009 on the control package pretty 
useless. And there's nothing I can do about it.
Even worse, this regression gets my thesis into trouble. As part of my thesis 
this spring, I have to (I'm happy to) publish a paper in a journal about 
control systems. The paper will demonstrate my new controller reduction and 
system identification functions - some of them are beyond the scope of Matlab - 
and propagate the use of GNU Octave in engineering.

Given the current circumstances, I would have to include remarks like "Due to a 
serious regression, which turned out to be a questionable feature, it is not 
possible to install the control package for the current Octave version 3.6. You 
have to use Octave 3.4, for which no good binaries are available."
Then the readers will think that GNU Octave is a ridiculous piece of 
alpha-quality software, far away from being a viable alternative to Matlab. And 
the bad thing is they would be right.

In order to circumvent these problems, I propose the following solution that 
works from both Octave prompt and OS shell:
Addition of a new option "--link-all-libs" to mkoctfile that brings back 
linking against all libraries from previous Octave versions (up to 3.4.3).

Best regards,
Lukas

[1]
http://savannah.gnu.org/bugs/?35114