<enebo[m]>
jeremyevans: yeah something with a command from rvm which downloads an empty version of itself. travis seemingly runs an extra rvm command which is not run for MRI
<headius[m]>
that's a combination of rvm and travis problems most likely
<enebo[m]>
AHA!
<headius[m]>
omg my build passed without the concurrency failure I've been looking at
<headius[m]>
and still running locally
<headius[m]>
a spoonful of synchronization helps the medicine go down
<enebo[m]>
kares: put out a new jdbc-sqlite3 on the 16th ... The jdbc-{sqlite2,postgresql} is the only visible change in displayed gem set
<jeremyevans>
enebo[m]: That would explain the error
<enebo[m]>
That though is weird since the jobs which were green was after the 16th
<enebo[m]>
hahahah no it is 18th
<enebo[m]>
ok so totally is the issue ... Locally I mst be locked to older version
<enebo[m]>
I will bundle update and see if I can get it happening local
<jeremyevans>
Sequel's last job ran 1 day ago, so it wasn't affected
<headius[m]>
sequel failed in my branch build also
<headius[m]>
which has nothing from master for days
<enebo[m]>
ok well progress. (not the database progress) so we will figure this out. I am really happy we run your suite in our runs :)
<headius[m]>
there are two remaining failures relating to concurrent autoload
<headius[m]>
this is going to be done before the weekend or I'll eat my hat
<enebo[m]>
jeremyevans: this is pretty interesting output and I can reproduce
<enebo[m]>
in older jdbc-sqlite3 it had like 739 runs or something around there and now it has 801
<jeremyevans>
I get the same 2 errors running with jdbc-sqlite3 3.27.2.1 on JRuby 9.2
<jeremyevans>
So it looks like not a JRuby bug, but a jdbc-sqlite3 bug. I'll update the Sequel travis config to not allow that version
<enebo[m]>
I also put in a puts for getBigDecimal call in jdbc.rb and before 3.27.1 I saw a handful and now I see tons of calls
<enebo[m]>
more runs probably means the db is advertising more features and running tests which have never run before
<headius[m]>
that's interesting
<enebo[m]>
I notice there are some 'end if DB.predicate?' stuff
<jeremyevans>
That's another possibility, I should probably look into that.
<headius[m]>
what were you fixing with the jdbc-sqlite release?
<enebo[m]>
kares released this but I believe it was to update the JDBC adapter for sqlite3 (which also will include sqlite3 impl) and postgresql
<enebo[m]>
so likely new sqlite3 probably may support more features and between our code and maybe sequel we are just hitting new behavior for the first time
<enebo[m]>
I can tell we hit more runs+assertions with this update in sequel doing: SEQUEL_SQLITE_URL="jdbc:sqlite::memory:" jruby spec/adapter_spec.rb sqlite
<headius[m]>
oh right
<jeremyevans>
If I had to guess, sqlite3 doesn't really support numeric/double types, so Sequel guesses based on column name, and one of the columns in the metadata query uses a column name that is interpreted as numeric/decimal but actually contains string
shellac has joined #jruby
shellac has quit [Quit: Computer has gone to sleep.]
<jeremyevans>
Fixed one of the bugs, about to look into the second one
<headius[m]>
Oh nice!
<jeremyevans>
Looks definitely like a bug in jdbc-sqlite3. It is using java.sql.types.NUMERIC for a field that does not contain numeric values for the PRAGMA foreign_key_list query
<jeremyevans>
Found the other issue. Similar issue, different pragma. That one is a regression because the pragma was supported in the old version
<jeremyevans>
yay, clean run in JRuby 9.2 with jdbc-sqlite3 3.27.2.1
<jeremyevans>
Let me run some more tests to see if the fixes break anything on other versions
<jeremyevans>
headius[m]: I just pushed fixes to Sequel, can you retry the JRuby Travis jobs and see if they work?