This project contains tools for running the perl6 spec test suite (roast) by several different implementations, as well as tracking the results of those runs.
See "perl6_pass_rates.csv" for the data, with a pretty version here: https://github.com/usev6/perl6-roast-data/blob/master/perl6\_pass\_rates.csv
Each implementation's last run is checked in under a corresponding ".out" file in the log/ directory. Check your implementation's file for any failures or unexpected passes.
Note - the percentage column shows the percentage of the implementations against each other for that day's run - so the highest number of passes is always 100%, and the others are in relation to that number. So, 100% doesn't mean "passes every spectest".
Each implementation shares roast with all the other implementations. If a test is added, it might not be immediately fudged to work on all implementations - in fact, depending on what is being tested, the entire file may start failing in some implementations.
The test numbers are run daily from a limited cron environment. If the test is expecting things (like an environment variable) these assumptions need to be added to the implementation's script in bin/
Finally, these tests are run on a schedule; it's not expected that commits to spec and to an implementation will be synchronized with the running of this harness; give new tests a day to settle in.
If you notice a test is continually failing (check the history of the implementation's .out file), please open a ticket with that implementation, or keep reading if you want to fix it.
The most complete implementation/vm combination, any failures here are from a regression, or from a newly added or changed test that doesn't pass. In either case, please open an RT (email to rakudobug at perl dot org), get a ticket number, and fudge the test, giving the ticket #.
There are currently two backends (MoarVM, JVM). MoarVM is run with HEAD and with the 6.c branch of roast. JVM is run with HEAD only.
A a test may fail on any or all of these backends. Be sure to mark the ticket in RT with the appropriate vm, and only fudge the test for the appropriate backend. (#?rakudo for all backends, #?rakudo.jvm for, e.g. the JVM backend)
These versions are in maintenance mode. Any tests that are not passing should be fudged with a todo if possible, or a skip if necessary.