Closed srbaker closed 9 years ago
I actually think the prefix should be dropped. Rubinius doesn't do it, and AFAIK neither does JRuby. The main reason I know for why we have it is because there are rake
tasks for MagLev, and one would need another Ruby to run rake
. That being said, I personally would like to see all the MagLev rake
tasks go away in favor of shell scripts, or something that doesn't require a whole `nother Ruby implementation to be installed.
Thoughts?
MagLev runs normal rake tasks just fine.
As Jesse pointed out, the problem is that rake scripts that involve starting/stopping/changing MagLev don't work because the MagLev executable has to be running in order for maglev-rake to work.
I looked into converting those tasks to shell scripts, but the complexity seemed high compared to simply using MRI or another ruby to run 'rake maglev:start' etc.
I had a shell script that renamed the binaries to avoid collisions or put them back. So I had the best of both worlds.
The benefit of leaving them unchanged is the ability to handle other scripts that use the canonical names. I had lots of those.
I think we should rather try to make the maglev-rake
unecessary.
As @timfel told me, rbenv or rvm may be are able to provide prefixed binaries for each vm incarnation. Probably we should focus on those.
Most of the names in
$MAGLEV_HOME/bin
conflict with system-wide installed defaults, preventing the user from using the system-wide defaults.MagLev's version of
ruby
,maglev-ruby
does the right thing and prefixes the bin withmaglev-
.Fix: All bins in
$MAGLEV_HOME/bin
should be named with amaglev-
prefix.