Currently we can see in the tree plenty of ebuilds needing to run "unset DISPLAY" manually to fix many different of issues (you can a few of them while grepping in the tree as comments in relevant ebuilds). Also, it's not the first time I see a test suite behaving differently depending on it finding a random DISPLAY and, then, for example, you seeing a window popping up suddenly while failing when running in console. Is there any reason for keeping DISPLAY set to a random value while building stuff? From my point of view, we should ensure packages are built without DISPLAY and, if an ebuild need a display, it should rely on virtualx.eclass to get the X stuff handled on a predictable way Thanks a lot
This would most likely go into section 11.1 "Defined Variables".
I don't mind this but then, I don't feel like adding more special variables is a good way forward. Maybe we should focus on the whitelist/blacklist option suggested elsewhere and WONTFIX/DUPE this one instead?
Indeed, this looks like a special case that the more general blacklist would handle. *** This bug has been marked as a duplicate of bug 499288 ***