You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
But that last part gets messed up if anybody does FEATURES=test USE=-test. That sounds odd, until you realize this is what USE=-* does :/
While of course we can say "Don't do that", we could also improve things if this flag wasn't defaulted to on globally. ( We could guard against this with RESTRICT="!test? ( test )", but doing that for hundreds of packages is pain ... especially when I'm using RESTRICT as the indicator I need to revisit the package to add tests in the first place :D )
This will especially matter when we're installing binary executables, we don't want the nature of those to be different simply because they were built by somebody who was wanting it tested.
Though the downside of that of course is, when we do ship binaries and build tests, the compile will have to do most of it twice, once for the binaries with -Cpanic=abort, and once for the tests ( without -Cpanic=abort ).
The text was updated successfully, but these errors were encountered:
kentfredric
added
the
discussion
ideas, thoughts and meta issues that may or may not eventually lead to some concrete change
label
Feb 10, 2020
Currently there's this in src_configure:
https://github.com/gyakovlev/rust-dev-overlay/blob/5bde43d1f07f3cf55ab899ff394635a830ae182d/eclass/rust-crate.eclass#L138-L148
But that last part gets messed up if anybody does
FEATURES=test USE=-test
. That sounds odd, until you realize this is whatUSE=-*
does :/While of course we can say "Don't do that", we could also improve things if this flag wasn't defaulted to on globally. ( We could guard against this with RESTRICT="!test? ( test )", but doing that for hundreds of packages is pain ... especially when I'm using RESTRICT as the indicator I need to revisit the package to add tests in the first place :D )
This will especially matter when we're installing binary executables, we don't want the nature of those to be different simply because they were built by somebody who was wanting it tested.
Though the downside of that of course is, when we do ship binaries and build tests, the compile will have to do most of it twice, once for the binaries with -Cpanic=abort, and once for the tests ( without -Cpanic=abort ).
The text was updated successfully, but these errors were encountered: