I'm a bit fuzzy on gem vs. stand-alone startup, but I don't think that
that is relevant here (since I see the same behavior in stand-alone.
In Stand-alone, we create a new runtime (using JavaEmbedUtils.initialize
after setting some very basic parameters), use loadservice.require to
bring in config/environment.rb, and our rack handler, then call
call(env) on whatever rails hands to our rack handler. We tried to stay
as hands-off with the Ruby codes as possible, to try to avoid weird
things like this.
Charles Oliver Nutter wrote:
Charles Oliver Nutter wrote:
Jacob Kessler wrote:
In the second (non-working) case, of Glassfish Gem,
RAILS_ROOT/app/models/subfo ends up in config.load_paths, the same
as with mongrel. $LOAD_PATH is slightly different, but it looks like
it's mostly related to not having loaded mongrel. However, when
rails needs to load post.rb, it fails with a NameError rather than
loading the file.
This seems to be a problem not directly related to loading. The load
process there is trying to load a file in order to define a constant
that does not exist. It seems like it's successfully loading the
file, but not producing the constant needed.
After some exploration, it appears that the load_path change in
environment.rb is getting clobbered at some point after it loads. I've
traced through the dependencies.rb handling of const_missing, and it
fails to locate the file because by that point the config.load_path
does not have the added path anymore.
Can you describe the sequence of events that happen during GF gem load
time? I suspect they're different than script/server, and as a result
the load path change is disappearing.
- Charlie
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
