> > I would like to replicate what nmake :PACKAGE: does without using
 > > nmake, since I am trying to cut down on build-system dependencies.
 > > This decision is partly at the behest of my colleague and partly
 > > because I continue to have difficulties getting bin/package to work
 > > the way I expect, which leads me to wonder whether I can distribute
 > > something based on AST that others will be able to build on systems
 > > different from mine.
 > 
 > you will get more leverage figuring out why "bin/package make"
 > fails to build nmake -- I and ast-users can help with that
 > 
 > one of the foundations of ast software (ksh first and then nmake)
 > is portability -- if it fails at then I want to know

I'm afraid that each time I approach bin/package I become overwhelmed
with its complex command-line API.  I last successfully built ksh from
source in 2006.  And although it's a petty irritation, I am continually
caught by surprise when help text emerges from stderr and not stdout.
I can't count how often I've typed 
   bin/package --help | less
in error.

 > > The ? is that I have gotten quite interested in understanding the
 > > details of how nmake works.  I am hoping to persuade the Glasgow
 > > Haskell Compiler project either to switch to nmake or to incorporate
 > > some of nmake's good ideas into their own build system.  Growing a
 > > formal model of nmake seems like a good next step, and :PACKAGE: is a
 > > not entirely random place to pull.
 > 
 > a convincing argument would be to set up the entire project with nmake
 > for all target architectures, culminating with a comparsion between the
 > nmake infrastructure against the status quo

Absolutely.  Ideally it's a big simplification for GHC.  Either way,
it would be a good learning experience.


Norman
_______________________________________________
ast-users mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-users

Reply via email to