On 10/17/10 14:34, Joel E. Denny wrote: > We could just extend the error message to explain the intent. It could > also explain that --no-bootstrap-sync is useful for trying out bootstrap > modifications. For longer experiments, it could suggest setting > bootstrap_sync=false in bootstrap.conf. Would that make it better?
Not much, I'm afraid. A common reaction will still be "I tried to bootstrap Bison and it didn't work. This software must be poorly maintained; I think I'll volunteer to work on something else." >> > For example, bootstrap could note >> > that it doesn't equal gnulib's bootstrap, and go ahead and use >> > the the gnulib bootstrap, all automatically. > We would likely overlook the note if it doesn't fail immediately > afterward. And the problem with that is....? How about another idea: have a very simple bootstrap script. All it does is fetch gnulib, and execute gnulib's more-complicated script. We could call our new script "boot", say. If this idea works, we could copy "boot" to gnulib as well, so that other projects can use the same idea. The "boot" script should rarely need to change, and the problems of it being out of date should be so rare that projects don't need to worry so much about keeping it up to date.
