On Sep 27, 2013, at 10:29 AM, Marcus Denker <marcus.denker@inria.fr> wrote:
On Sep 27, 2013, at 10:08 AM, Marcus Denker <marcus.denker@inria.fr> wrote:
On Sep 27, 2013, at 9:25 AM, Pavel Krivanek <pavel.krivanek@gmail.com> wrote:
And it's interesting that during the first pass on a dirty image it works well even on the slave too.
I will revert the compiler merge for now
(it did even run the regression tests⦠but showing some strange - I think - trait-recompiling related bug..)
So what we will do
-> revert this update (it is already auto-reverted due to the crash) -> save Opal as it is in the image before this to to the Opal repo -> get the Opal regression green (traits bug) -> get image green (we do have some failing tests) -> *then* on the Opal repo, merge step-by-the intermediate 5 small commits (Abstract compiler class, some trivial cleanups, clean block analysis that is not called) I do not see how these changes could lead to the problem, but it could be that there is some bug in the compiler that shows itself when recompiling certain methods.
One interesting observation: we did not check the Opal-Regression tester for 4 weeks, and we are in the mess. This just shows how important continous integration + automatic regression testing *after every commit* is for compiler development⦠(or being more intelligent would maybe another solution ;-)
Result: it still fails, even after reverting.
Are we sure that the VM did not change in the middle? It looks *really* strange!
https://ci.inria.fr/pharo/job/Pharo-3.0-Update-Step-2.1-Validation/label=lin...
It seems someone changed the config to load the latest VM which has problems right now. I will revert that. Marcus