Default option isn't feasible due to SandyBridge regression. To keep binary package users happy let's flavorize.
Diff Detail
- Lint
No Lint Coverage - Unit
No Test Coverage - Build Status
Buildable 20856 Build 20235: arc lint + arc unit
Event Timeline
Apart from the SUFFIX issue I'm ok with this.
multimedia/libva-intel-driver/Makefile | ||
---|---|---|
39 | While @ is used to differentiate flavors though origin for ports, usually for PKGNAME we're using - as separator. Personally I'd prefer PKGNAMESUFFIX=-${FLAVOR} which is more in line with what other ports are doing. |
- Add CONFLICTS_INSTALL (doesn't work for some reason)
- Adjust pkg-descr to document VP8 encode and VP9 decode target different platforms
- Pacify "FLAVORS" has to appear earlier from portlint
multimedia/libva-intel-driver/Makefile | ||
---|---|---|
39 | BTW it could also be an helper: hybrid_PKGNAMESUFFIX=-hybrid |
multimedia/libva-intel-driver/Makefile | ||
---|---|---|
39 | If I switch to - as separator then users may confuse libva-intel-hybrid-driver (different package) with libva-intel-driver-hybrid (flavor). - is probably more common in the ports tree because many slave ports used it, so during slave->flavor conversion maintainers chose to keep package name compatibility. This port never had slaves, so compatibility isn't a concern. |
*_CONFLICTS_INSTALL lines appear to be garbage. I've tried in this port and security/openssh-portable but make install isn't interrupted early unlike with slave ports. portmgr, did CONFLICTS* machinery ever work with flavors?
multimedia/libva-intel-driver/Makefile | ||
---|---|---|
39 | Please understand I'm expressing this not as a comment against you. I don't really care if we use a - or @ here, so I'll approve this review whatever the separator used. I'm anyway taking the chance to state that we need a consistent way of doing this in the ports tree, also not to confuse users. My opinion is we need an official way to handle this, which could also be - for migrated slave ports and @ n other cases, |
portmgr review is too opaque i.e., the reviewer fails to provide rationale for why things have to be done certain way. If this is to be expected to get FLAVORS into ports may as well go back to slave ports.
D18017#385299 is also yet to be answered.
--enable-hybrid-codec can only be implemented as conflicting package i.e., either a slave port or a flavor. During runtime i965_drv_video.so loads hybrid_drv_video.so provided by the dependency i.e., hybrid can only be used through i965 driver. For an alternative see libva-intel-media-driver which also have hybrid mode
multimedia/libva-intel-driver/Makefile | ||
---|---|---|
23 | Why? @* makes default flavor conflict with any other flavor. | |
37 | Ideally, it should be default_PKGNAMESUFFIX_OFF=@${FLAVOR} but _OFF helper is harder to implement without ugly .if conditionals that break declarative style. Where is it documented that only - is allowed as a separator for PKGNAMESUFFIX/PKGNAMEPREFIX? Or did you miss that by using - it'd be easy to confuse libva-intel-hybrid-driver with libva-intel-driver-hybrid? |
multimedia/libva-intel-driver/Makefile | ||
---|---|---|
37 | -compiled.specifics convention predates flavors but poudriere uses @flavor when building them thus breaking discoverability rationale. If the convention is inconsistent it can be interpreted the same as no convention. Otherwise, PKGNAMESUFFIX=.hybrid can be used instead as -compiled.specifics already defines . (dot) as allowed character. |
Obsolete after 0fe260b63e82. I'd like to avoid flavors for what should work fine together.