On 14 Jun 2017, at 09:39, Marcus Denker <marcus.denker@inria.fr> wrote:
On 13 Jun 2017, at 13:44, Holger Freyther <holger@freyther.de> wrote:
Hi!
today I started to look at of why my benchmark in the CI stalled/didn't complete and I see three separate issues and wonder how to address them properly? I have a truncated log attached
Issues: * In my CI script I disable Deprecation warnings but this has no influence on Deprecation>>#transform. This means that it will try to write the changes?! To scale we start the same image multiple times.
* Somehow writing the log fails (and the TCP connection fails the same time?!) with a primitive error.
* Errors "recurse". So while we try to exit the image it seems to fail the same way and tries again.. which fails again.. which fails again..at some point no new context will be allocated.
Solutions?
* Prevent new errors in NonInteractiveUIManager>>#unhandledErrorDefaultAction:?
* Check in >>#transform if transformation is disabled and then return early?
yes! This should have been like that.
* ??? No idea.. I will strace to see what happens to the Debug log.. but ideally I would like this file writing to be disabled. :)
Should I create tickets?
Yes! For the Deprecation one I can prepare the fixes for Pharo6 and 7. The other I have no idea yet, but having it on the tracker would be good.
I added https://pharo.fogbugz.com/f/cases/20148/transforming-deprecations-should-tak...