Remaining differences between modular & monolith installs

Matthieu Herrb matthieu.herrb at
Mon Oct 17 12:34:33 PDT 2005

Alan Coopersmith wrote:
> Matthieu Herrb wrote:
>> Alan Coopersmith wrote:
>>> I've been diff'ing install trees produced by the modular & monolith
>>> CVS head's lately.   (You may have noticed all the man page checkins
>>> to bring the modular man pages up to what's installed by the monolith.
>>> Another set should be coming soon for the driver man page.)
>> Unfortunatly those changes break the build on BSD systems in two ways:
>> - the .shadows.DONE: target confuses make; it interprets it as an 
>> implicit rule (because automake adds .SUFFIX declaration automagically).
>> - the LIB_MANSUFFIXvariable is not expanded in constructs like 
>> $(Xau_shadows:=.$(LIB_MAN_SUFFIX)). It can be forced by using an 
>> intermediate variable affected with the ':=' operator.
>> The attached patch shows what is needed for libXau. Unfortunatly I 
>> don't have time (and it would be ugly) to do the corresponding changes 
>> to all modules. So if someone has a better idea to implement this in a 
>> portable way, it's welcome.
> Would it be simpler and more portable to replace $(LIB_MAN_SUFFIX) with
> @LIB_MAN_SUFFIX@ so automake replaces it before make sees it?

yes it does.
This and using shadows.DONE instead of .shadows.DONE fixes this issue.

Note that I still haven't finished getting a working server to build on 
OpenBSD, mostly by lack of time, not because of major problems.
Matthieu Herrb

More information about the xorg-modular mailing list