# $Id: TODO,v 1.31 2012/08/12 17:10:42 ksb Exp $ I had a strong intuition that an environment transistor for the dashdash token that was a dicer expression on the HXMD_LIB value would be more than a little useful in the future, but I can't remember why. The default would be "%[1:2]" where 1 is HXMD_LIB, and 2 was the built-in default -- or something like that. I think we'd read the value from $HXMD_DASHDASH or so. That would force us to include the dicer.c file in mmsrc, Oh, save me. Then we could implement xapply in mmsrc... no. Could be the rule for double-dash should be first absolute path that exists. Not just the first one. Maybe. Under -d M (debug missing stuff in config reader) we should vetch about MACRO=naked-value since distrib will fail for the missing quotes (which hxmd doesn't really need). That would be in hostdb.m about about line 500, after the case for HT_WORD and HT_MACRO, we'd check for the debug flag and carp there. We'd also want to only carp once per configuration file, for sure. The cache directory location (in .) forces some msrcmux pulls to fail because they cannot write on the cache directory. We should overlay a gtfw on the cache. Then code the Cache.m4 and such to use the gtfw overlay when one is in-scope. And msrcmux would need to be wrapped in a gtfw, which is easy. Nothing else. Make a book or better website for the use of this tool-chain. ---- Other ideas ---- Allow an "export" command for macros to push them into the environment for the subprocess. The smacks of the export GNU make does that breaks all sorts of things. I think you can put the export in if you want it. We've thought about these, just so you know we didn't just put in every idea that came to us (mostly). I turned down -Z and -C allowing "/dev/null" as a specification. Today they reject special files, because it is mostly silly. We document that the file "empty" in .../lib/hxmd is an empty file, use that (-Cempty). If you want a FIFO use ... -C -