The jar loading logic is done in LoadService, and was a half-assed attempt for us to make .jar files load something like .so files (load and initialize at once). We need a better mechanism (probably based on meta-inf) but I have been manually loading libraries via JRuby.runtime to move away from the old mechanism.
- Charlie On Thu, Jan 10, 2013 at 12:45 PM, Matt Hauck <mattha...@gmail.com> wrote: > Turns out OpenSSLReal.createOpenSSL is getting called twice. It was called > once by the OSSLLibrary.new.load call, but it was apparently called also > when jopenssl.jar was loaded. I cannot figure out why it is getting called > when the jar is loaded... > > -- > Matt Hauck > > On Thursday, January 10, 2013 at 3:35 AM, Matt Hauck wrote: > > Okay. Stayed up a bit too late for this… but finally got all the commits and > such sorted out: https://github.com/matthauck/jruby/commits/bc147 > > I'm strangely getting logs of "already initialized constant" errors now > though coming from this line: > org.jruby.ext.openssl.OSSLLibrary.new.load(JRuby.runtime, false). Not sure > what's going on there... > > -- > Matt Hauck > > On Wednesday, January 9, 2013 at 10:44 PM, Matt Hauck wrote: > > I've almost got it working now. There was a regression in > Signed#digestAlgorithmsToASN1Set. Getting closer… > > -- > Matt Hauck > > On Wednesday, January 9, 2013 at 10:40 PM, Lars Westergren wrote: > > Hmm, interesting. I'll also look into it this weekend, if you haven't solved > it by then. Good thing you spotted this. > > On Jan 10, 2013 3:24 AM, "Matt Hauck" <mattha...@gmail.com> wrote: > > FYI: I'm continuing to work on this, but just an FYI there are definitely > some problems with the pkcs7 portions of the library. I've attached a simple > script that is breaking with the new 1.47 port. It thinks it signs the data > successfully and all, but the to_pem / to_der generates invalid data. The > script tries to open the generated signed.pem with openssl and openssl can't > read it. This script passes fine with jruby 1.7.1. > > -- > Matt Hauck > > On Wednesday, January 9, 2013 at 11:58 AM, Matt Hauck wrote: > > Hmm, the problem appears to be that the actual tests have changed. One of > the last tests failing after merging Lars' BC 1.47 changes is in > test_pkey_rsa.rb, which is entirely different on the jruby repo from what it > used to be in the jruby-openssl repo. The test is "test_load_pkey_rsa_enc". > The jruby repo's copy of that file doesn't even have any PEM loading tests… > > It looks like something got messed up (perhaps intentionally?) with the > openssl tests when they were merged to the main jruby repo > > -- > Matt Hauck > > On Wednesday, January 9, 2013 at 10:07 AM, Matt Hauck wrote: > > Sorry for no context, I have a private fork of the old jruby-openssl gem > (forked from 0.7.5) that we are still using internally due to some fixes > that hadn't yet been accepted upstream. So I merged Lars' BC 1.47 changes > back there and ran it against those test cases and got some failures, though > running test:mri19 does not indicate any failures... > > -- > Matt Hauck > > On Wednesday, January 9, 2013 at 9:52 AM, Charles Oliver Nutter wrote: > > I'm not sure what you mean by the "old" openssl gem. If you run rake > test:mri19 in the JRuby repo it will run the 1.9 version of the > OpenSSL tests with known failures excluded. > > - Charlie > > On Wed, Jan 9, 2013 at 11:12 AM, Matt Hauck <mattha...@gmail.com> wrote: > > Hmm. I don't know if all the tests are passing. It is possible I am doing > something wrong. I copied over the source to the old jruby openssl gem > (since that's the only place I knew how to run the openssl tests), and some > of them were failing. The biggest change i needed to get them closer to > passing was in Recipient#_initialize. > > How do you run the openssl tests only from the jruby repo? > > -- > Matt Hauck > > On Wednesday, January 9, 2013 at 8:42 AM, Charles Oliver Nutter wrote: > > Excellent, Lars! I'll take a look at it and probably roll it into > master if it runs green. It will be *very* nice to be current on > Bouncy Castle. > > - Charlie > > On Wed, Jan 9, 2013 at 1:40 AM, Lars Westergren > <lars.westerg...@gmail.com> wrote: > > I'm far from an expert on security, so if anyone is, a quick code inspect of > the above commit would be appreciated! All the build tests pass now, but I > haven't had time to try it out on a real application. > > Regards, > Lars > > > On Wed, Jan 9, 2013 at 6:14 AM, Matt Hauck <mattha...@gmail.com> wrote: > > > Sweet. Perfect timing. It looks like it's pretty far along! I'm glad > people that know more about this stuff are already working on it. =) > > -- > Matt Hauck > > On Tuesday, January 8, 2013 at 9:05 PM, Alex Tambellini wrote: > > The upgrade to 1.47 is already being worked on here: > > > https://github.com/LarsWestergren/jruby/commit/c9f36d616bebe5a96c93908469d974c64de314fe > > On Jan 8, 2013, at 11:10 PM, Matt Hauck <mattha...@gmail.com> wrote: > > I am glad to hear that I will at least be able to upgrade to jruby 1.7.2 > and replace out the built-in openssl support in case I need to make a fork > with this bouncy castle 1.47 support. I've got it compiling now with the > updated BC now, and am beginning to work through some more the more > difficult changes. The biggest has to do with deprecating X509Name in favor > of X500Name and the various implications of this. > > -- > Matt Hauck > > On Tuesday, January 8, 2013 at 8:00 PM, Matt Hauck wrote: > > Sadly not so. If you try replacing the bc*.jar files in build_lib with > bcprov and bcpkik 1.47 versions you will find the build fails with 100 > errors. > > -- > Matt Hauck > > On Tuesday, January 8, 2013 at 7:01 PM, kristian wrote: > > Jruby 1.7.x comes with openssl bundled. I think from version 1.7.1 onwards > bouncy castle gets a java package rewrite and is used only internally for > openssl. So you are free to add any version of bouncy castle as needed by > your application - just do not add the jruby-openssl since this will pull in > the old version. > > Kristian > > > --------------------------------------------------------------------- > 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 > > > > > > > --------------------------------------------------------------------- > 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
