At 09:29 18/03/05 +0000, Simon Peyton-Jones wrote:
[Redirecting to GHC users, which is where GHC-specific questions belong]
OK (but I'm not subscribed to that list, so pls copy me on any further comment, if any.)
| OK, it's not so bad. If I delete the AffyDataCalc.o file before compiling | AffyDataCalcTest, it all works fine. ... | Hmmm... is there something the GHC --make feature can do to force | recompilation of the other modules previously compiled as main | modules? Just a thought.
We're always wary of adding too much magic!
I can appreciate that. It did strike me as a little strange (magic, even?) that the result of compiling a module named explicitly on the GHC command line is different from the implicit compilation performed when chasing imports. (I think I understand *why* it happens, but it still seems a little strange.) By way of exploring the issue, rather than actually asking for any of the following to be done, some thoughts I had are: (a) generate the program initialization/finalization entry points in a separate .o file from the compiled module, and include that in the link. (b) use a different output filename (e.g. module_main.o) when compiling a module as a program entry point. (c) combination of (a) and (b) - compile base module as non-main, and a separate output file for the program startup magic with a name based on but different from the corresponding source module.
I'll add a note to the -main-is documentation though.
Thanks. #g ------------ Graham Klyne For email: http://www.ninebynine.org/#Contact