Working for a large company, we are using the parallel_tests gem to run our cucumber automation. This works well for our Ruby divisions, but for our Jruby folk we need another option. The issue is that Jruby (as of 1.7) fork spawns a new java JVM without the JVM options. I have figured out that I need 1 of 2 solutions and I'm not exactly sure how to implement either of them.
We need to be able to pass the JVM options into fork so we can get all the options we need.
or
We need to change how the gem handles forking so the processes all run on the same JVM
I don't know if either of these solutions are a possibility, but maybe someone else will know better
According to a solution posted at Alternative for spawning a process with 'fork' in jRuby?
I found out the solution for this. We can use the built-in library FFI in JRuby to 'simulate' the Process.fork in MRI.
To mimic the Process.fork in MRI Ruby
module JRubyProcess
require 'ffi'
extend FFI::Library
ffi_lib FFI::Library::LIBC
attach_function :fork, [], :int
end
pid = JRubyProcess.fork do
#do_your_work
end
More details:
https://github.com/ffi/ffi
http://blog.headius.com/2008/10/ffi-for-ruby-now-available.html
another option was to use spoon:
https://github.com/headius/spoon
I am no linux guru, but these seem like simple ways to implement Jruby. Does anyone know if this is an effective solution?
I don't think there is any safe way you can fork the jvm like you can ruby. There is a gem called childprocess that provides cross platform process managemant.
I've created a gem called jcukeforker that can run both features and scenarios in parrellel in jruby.
Feel free to submit an issue or a pull request for setting environment variables on the subprocesses. This is only relevant if you want the subprocess to have different environment variables than the parent process.
Related
I'm trying to compile some raku code I saw on https://replit.com/languages/raku. The code is from Why is Raku reporting "two terms in a row" when I define a new operator?.
It begins like this:
unit module Format;
use List::Util;
...
It fails to compile with:
raku ./main.raku
===SORRY!=== Error while compiling /home/runner/l4gp3hvdnhd/./main.raku
Could not find List::Util in:
inst#/home/runner/.raku
inst#/opt/rakudo-pkg/share/perl6/site
inst#/opt/rakudo-pkg/share/perl6/vendor
inst#/opt/rakudo-pkg/share/perl6/core
ap#
nqp#
perl5#
at /home/runner/l4gp3hvdnhd/./main.raku:3
exit status 1
On the other hand I see this is a valid module - https://raku.land/zef:lizmat/List::Util.
Why is it failing?
TL;DR Run zef install --/test List::Util in the console, put use lib '.'; at the top of your Raku main.raku, and run, don't walk, with your program, before gremlins gleefully render your efforts in vain. Or maybe just listen to Liz and Rawley.
As Liz and Rawley have noted, you need List::Util installed.
But while I largely agree with them in practice (it may be a pain to use replit to do what you're trying to do) I think a different response to complement theirs might be helpful.
One of the ways replit is trying to distinguish itself from other online evaluators, is that it is trying to be akin to a full dev environment.
In reality it's early days in their ambitious project, and beggars can't be choosers (if you're not paying, it's hard to complain if things don't work out as you might want), but of particular relevance for this SO it is worth noting that it does have console/shell facilities and they've installed Rakudo Star, or perhaps just something like it, including the Raku package manager pretty much everyone uses (zef).
Thus this command, which I just ran in replit's console of a new raku session, worked:
zef install --/test List::Util;
(The --/test tells zef not to run tests. I've only got a free account and it looked like replit killed zef's process when I ran just zef install List::Util during its running of tests. Presumably they take too long, but I don't know.)
And then this main.raku also worked:
use lib '.'; # Tell Raku(do) libs are in current directory.
use List::Util <notall>; # Load and import `notall` from module.
say notall { 42 }, 99; # Try it.
But now the rub. As I was composing this answer, the expected happened. My internet connection momentarily flaked out, the replit rebooted the session, and while my main.raku code was rescued, both List::Util and my console history had disappeared, so I had to paste the install command again and rerun it to get the module installed again.
It's all just throwaway container magic, and there's only so much replit has done thus far to make the simulation of a real full local dev environment really work.
Maybe if your Internet connection is rock solid and/or you're using a paid replit account and/or it's the full moon, it'll all work out. Or maybe you're best off following Rawley's advice.
Speaking of which (I mean Rawley's advice to set up your Raku dev environment locally), if you do install locally you can also install the awesome free version of CommaIDE.
You do not have List::Util installed. Since you're using an online interpreter you will most likely have a lot of trouble doing this. Instead I recommend installing Raku on your local machine with rakubrew.
Then run the following commands:
rakubrew build # Make sure to follow the instructions at the end
rakubrew build-zef
zef install List::Util
Now you should be able to run your code on your local machine, and you'll have access to the List::Util library.
Does anyone know how to use the make proxy command from the command line in order to see the documentations of the modules? I have installed Instant Veins 5.0 i2 with SimuLTE. It doesn't work generating from the IDE of OMNeT++, it returns an error and the cars simulation can't run, so i need to re-install the whole instant-veins-5.0-i2-simulte.ova file from scratch.
For "Veins LTE 1.3", module documentation can be generated by opening a terminal, changing to the veins/ directory (the one that contains sumo-launchd.py), and running make doxy. Module documentation is then available in doc/doxy/index.html.
Note, however, that Veins LTE has not been in active development for quite some time and integrating much more recent versions of SimuLTE and Veins is now possible via Veins_INET. For more details, see https://veins.car2x.org/documentation/modules/#lte
I would like to run a local Monticello HTTP repository at work, so that we can share code easily among colleagues.
Is there a way to run something similar to SmalltalkHub privately?
EDIT:
I have tried all the options here and neither of them seems to work smoothly. Let me recap the options:
1) WebDAV on Apache, following Stuart. I have tried it, following some online guides. My current dav.conf looks like this:
DavLockDB /tmp/DavLock
Alias /pharo /opt/data/pharo
<Location /pharo>
Order Allow,Deny
Allow from all
Options Indexes MultiViews
Dav On
AuthType None
</Location>
I worked for a few days. Then suddenly I am not able to read new versions of a certain package. Whenever I write a version in an image and read it in another one, I get an exception ZnInvalidUTF8. I am not sure why, it may be that WebDAV has issues listing too many files?
2) Setting up my FTP. It seems to work, but when I try to set this repository as a remote in the versionner I get MCFtpRepository doesNotUnderstnd: #koRemote
3) SqueakSource3, following Tobias. I have tried running the two Gofer commands in both Pharo2 and Pharo3. In Pharo2 it does not load at all. In Pharo3, more or less everything works. I had to fix a few errors due to deprecated or removed messages, but in the end I am able to create projects and write to them.
The problem arises when I read. Apparently SS3 keeps some kind of internal cache. The result is that the list of packages I see on the project page is different from the list of packages that the client gets. The difference seems to be that the client requires a short version of the page, like http://localhost:8080/ss/MyProject/?C=M;O%3DD, and the results there are consistently less than in the full page http://localhost:8080/ss/MyProject.
Moreover, even on the project page, the list of versions remains cached until I navigate on a different project.
4) SmallTalkHub, following Sean. I have tried both using images from the INRIA server and images suggested from the Pharo-VM-loader (they may be the same).
I had to install Seaside again, since there was no ZnZincAdaptor in the downloaded image. I am now able to start SmallTalkHub, but as soon as I try to register a user, I get an error MessageNotUnderstood: receiver of "new" is nil. I am not able to track where this error comes from (is there a way to open a server-side debugger instead of returing 500 in Seaside?).
After this error, I can see a user both in mongodb and in the interface, but I am not able to login.
5) Git using filetree, as suggested by Kylon. This would prevent me from using MetaCello to handle dependencies and looks even more compelx than the other options.
At this point I am at a loss. :-( If I want to use Pharo, I will need to be able to collaborate with my colleagues. Using open source repositories is not an option, at least right now.
Is there a simple, tried and tested way to set up such a repository?
SqueakSource3 or SmallTalkHub would be even better, thanks to their user interfaces, but I really need at least basic collboration. Having an option that can run on a headless server would also be a big plus, as if this becomes a tool we use, it will not be doable to host the repository on my laptop.
Per this thread on the Pharo Dev mailing list:
Setting up the Server:
Download a SmalltalkHub image (https://ci.inria.fr/pharo-contribution/job/SmalltalkHub/)
Install mongodb on your computer (for Debian: apt-get install mongodb)
Launch the SmalltalkHub image
Evaluate: ZnZincServerAdaptor startOn: 8080
Visit http://localhost:8080/tools/hub, create an account and a project
In addition to Sean's answer - if you just want a Metacello repository, and don't necessarily need the full SmalltalkHub stuff, then you just need a WebDav server. Apache will work fine, and I've even used Confluence's WebDAV support (with some tweaking) successfully in the past.
In addition to the other answers:
Just storing your versions in DropBox work very well!
You can also install SqueakSource3 (like SmalltalkHub, doesn't need MongoDB):
Gofer new
url:'http://www.smalltalkhub.com/mc/Seaside/MetacelloConfigurations/main';
package: 'ConfigurationOfSeaside3';
load.
((Smalltalk at: #ConfigurationOfSeaside3) project version: #stable) load.
Gofer new
url:'http://www.squeaksource.com/MetacelloRepository';
package: 'ConfigurationOfSqueakSource';
load.
((Smalltalk at: #ConfigurationOfSqueakSource) project version: #bleedingEdge) load: #('All').
Then start your Adaptor (eg ZnZincServerAdaptor startOn: 8080) and goto http://localhost:8080/instalSS
Another way is go down the popular route of Git. I am using Github for my projects and it works great while Git itself works very well locally too. So if are already familiar with Git then its a very good choice
You can find more information here https://ci.inria.fr/pharo-contribution/job/PharoForTheEnterprise/lastSuccessfulBuild/artifact/GitAndPharo/GitAndPharo.pier.html
Sorry about the bad smalltalkhub experience. I have made some fixes to the configuration, and need to check if that is enough
I am using Selenium locally to run two .feature files. Both pass locally. Neither of the tests interact with a db (nor will future tests). I would like to use the parallel_tests gem to spin up two selenium web browsers concurrently and run each .feature file. I have tried to follow the readme on gems homepage, but am still having no luck.
I can run rake parallel:features and I get the following output:
Using recorded test runtime
8 processes for 2 features, ~ 0 features per process
However, it then proceeds to fail immediately and informs me that I have not defined the scenarios.
I am using Rails 3.2 and Capybara
I have also tried adding begin; require 'parallel_tests/tasks'; rescue LoadError; end
to the top of my Rakefile which I found elsewhere, but doesn't help.
Try using:
parallel_cucumber -n 2 features
I currently run my Rails app using:
jruby --1.9 -J-XX:+CMSClassUnloadingEnabled -J-XX:+UseConcMarkSweepGC -J-XX:MaxPermSize=256m -S rails server
This is getting pretty old now. How can I set my Rails project up so that just running
rails server
has the same effect?
(Note: bash aliases and the like are not what I'm looking for here. I want to make the project work right, not fix my local settings)
When using RVM and a project .rvmrc, the canonical way is to set PROJECT_JRUBY_OPTS in the project .rvmrc. A bug prevented this from working for me, so use rvm head.
If not using rvm then use JRUBY_OPTS, which is the built-in way of doing it that JRuby checks (in fact, the PROJECT_JRUBY_OPTS thing ends up being converted to JRUBY_OPTS by rvm).