In principle MRI is always right, so unless there's a comment or spec that explains why we deviate it's a bug. Eloy On 24 aug. 2011, at 18:19, Steve Clarke wrote:
I want to use PostgreSQL with macruby.
At first I couldn't make this work. I eventually had some success and learned quite a lot on the way. I'm not sure whether I found a bug in macruby but it certainly behaves differently from ruby 1.9.2-p290 installed from macports. The latter worked with postgres with no problems.
My first try was with macruby 0.10 stable. I couldn't get the pg gem (pg-0.11.0) to compile and eventually gave up. There were lots of obscure errors which I regret I've not recorded.
I then moved onto macruby 0.11 which I compiled from source that I checked out today. pg-0.11.0 now compiled but when I tried to use postgres from my macruby test I aborted with a failed dynamic link to ENCODING_GET_INLINED. I soon realised this was another name for ENCODING_GET and I added the definition to macruby's encoding.h.
I could now get as far as connecting to postgres and retrieving a result set. However, as son as I tried to examine any of the values in the result set I aborted:
Assertion failed: (idx >= 0 && idx < ENCODINGS_COUNT), function rb_enc_associate_index, file encoding.c, line 590. Abort trap: 6
I eventually found that pg was passing an idx of -1. I modified the the pg code to prevent this happening and it now works fine, as far as I've gone.
It seems to me that pg should not pass an index of -1. ruby 1.9.2-p290 however seems to accept it. So who's right ?
Hope this is of interest to someone.
Steve
_______________________________________________ MacRuby-devel mailing list MacRuby-devel@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel