Re: Make and makefile bugs (PRs 49085, 49086, 49087)

[christos%astron.com@localhost (Christos Zoulas)]

In article <20140817150321.GT965%roskakori.fi@localhost>,
Jarmo Jaakkola  <jarmo.jaakkola%roskakori.fi@localhost> wrote:
>There's one thing about the suffix transformation rules I forgot
>to bring up earlier.  It only comes up if one suffix is a catenation
>of two other ones.  The current implementation always gives precedence
>to the two suffix rule and this is how I left it.  I was just wondering
>if it is the best way.
>
>So, if .a, .b, and .a.b are known suffixes, what should a rule for .a.b
>mean?  If two suffix rules always take precedence, there is no way
>to have it mean "foo.a.b" -> "foo".  Should it perhaps be that .SUFFIXES
>order is used here too?  So if the order is for example ".b .a.b .a",
>then it would be a single suffix rule, but with ".a .a.b .b" it would be
>a two suffix rule.  That way the user could decide instead of make doing
>the decision.
>
>It might be possible to implement it as both simultaneously, but I think
>that would just get complicated.

This is the problem with suffix rules and why pattern rules are better...
I had this issue many years ago (when foolishly tried to use filenames
with multi-level suffixes), and I gave up.

christos