Fix \mess not being able to accept some commands

[ElectronicRU]

The newline between \node and node text in \messanother prevented TikZ parsing tricks that suspend command evalutation from working.

In my use case, \ref inside \mess broke horribly,while working fine in messcall and call.

[hmenke]

Can you provide a minimal example for the failure because as far as I can see you have not made a semantic change.

[ElectronicRU]

Ah sure - I was using \lstinline (from lstlisting) in the \messanother and it kept breaking horribly until I made this change. Calls didn't break, that's how I divined what the difference is.

[hmenke]

You cannot use any verbatim material in \mess and your patch doesn't fix that.

[ElectronicRU]

Well, I was able to... and in \call's the verbatim material works fine as well. You're welcome to go ahead and try.
TikZ does have some delayed-evaluation thing that makes it work, but it only works if there's no newline there.

[hmenke]

You're welcome to go ahead and try.

Sure thing, here you go! This doesn't work, neither with nor without your patch.

\documentclass{article}
\usepackage{pgf-umlsd}
\usepackage{listings}
\begin{document}
\begin{sequencediagram}
  \newthread{a}{:Thread A}
  \newthread[gray]{b}{:Thread B}
  \mess[1]{a}{\lstinline|{|}{b}
  \mess[1]{b}{\lstinline|#|}{a}
\end{sequencediagram}
\end{document}

TikZ does have some delayed-evaluation thing that makes it work

There is no such thing.

[ElectronicRU]

Oh you're right, my memory is hazy. The patch indeed didn't fix \lstinline, but it did fix \ref and \autoref usage in messages.
I apologize, it's been half a year and I mixed things up.

(And I even written it in the MR message and still got it wrong!)

[hmenke]

No, \ref and \autoref are not fully expandable so it will still blow up when used as the node name.

pgf-umlsd/tex/latex/pgf-umlsd/pgf-umlsd.sty

Lines 269 to 270 in 8766cc1

	\node (#3 from) at (mess from) {};
	\node (#3 to) at (mess to) {};

This library just has to be rewritten (like I already did for pgf-pie).