Nam Nguyen <namn@messagemode2.com> writes:

> Greg Steuck writes:
>
> Thanks for this. First, some feedback on the port itself:
>
> 1) Should this duplicate MODCABAL_VERSION be here?
> MODCABAL_VERSION =      0.47.1
> duplicated assignment of MODCABAL_STEM at Makefile:27
> duplicated assignment of MODCABAL_VERSION at Makefile:28
> duplicated assignment of MODCABAL_MANIFEST at Makefile:29

Oops, this is bug.

> 2) Is inotify used in WANTLIB?
> xmobar-0.50(x11/xmobar):
> Bogus WANTLIB: inotify.4 (/usr/local/bin/xmobar) (NOT REACHABLE)

Yeah, as Stuart said it's a tooling limitation:

% ldd /usr/local/bin/xmobar | grep libinot
	000000024c63e000 000000024c649000 rlib  0    1   0      /usr/local/lib/inotify/libinotify.so.4.0

> $ ./build xmonad 
...
> cbits-unix/init.c:3:10: error:
>      fatal error: 'sys/random.h' file not found
>         3 | #include <sys/random.h>
>           |          ^~~~~~~~~~~~~~
>   |
> 3 | #include <sys/random.h>
>   |          ^

Sadly, I've seen this too. The original problem is the upstream bug.
https://github.com/haskellari/splitmix/issues/101

A workaround is to use `cabal freeze` then change cabal.project.freeze
to avoid the bad version. This makes the build hermetic and prevents
cabal from picking up the broken version.

Thanks
Greg