Re: function $(eval ...) in make 3.81 compared to 3.80

[Jacques Klein]

Thank's Philip G. for replying, here my comments:

First, I added the $(eval $(warning ....)) stuf just to convince myseft 
about how $$ is "evaluated", and for that mater, like you say,  the 
outer $(eval is just a "no-op", ( it's my way-of-doing to debug $(eval 
stufffff). i.e check for correctness of first and second expansion rounds).

And you gave me the "correct hint" in order to understand the problem, 
"the semicolon bug" in 3.80, I "solved" some years ago with
backslashing.
Your solution which is to put the complete "target: ; command" line into 
a var works fine,
and I tested also the following one:

%sc := ;
$(eval target: $(%sc) command)

which is to just put the semicolon into a var (of course not necessairy 
with 3.81).

Finally, I think I prefer your solution, since it allows to write the 
complete rule in the "usual" way (multi-line with <TAB>command) by using 
define.

But most important is:   the "double expansion" is definitivelly the 
"right way to reason" with $(eval ...)'s.