| Summary: | Make java eclasses warn when there's both virtual/jdk and virtual/jre atoms in DEPEND | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Petteri Räty (RETIRED) <betelgeuse> |
| Component: | [OLD] Java | Assignee: | Java team <java> |
| Status: | CONFIRMED --- | ||
| Severity: | normal | CC: | asturm |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
repoman actually spits out a dependency.missingslot warning if you get this wrong but the error message doesn't relate specifically to this problem. All our >= virtual entries fall foul of dependency.missingslot but repoman knows that this is okay for DEPEND=">=virtual/jdk-x.x" and RDEPEND=">=virtual/jre-x.x". It's only when you mix them up that it complains. I know to look out for this but other developers have misinterpreted it and incorrect restricted the virtuals to a specific SLOT. In rare situations, such as ant-core, it is correct to mix these up so it shouldn't be a fatal error. By the same token, it should probably be a repoman warning rather than an eqawarn. |
16:16 <@Betelgeuse> ssuominen: then you don't want virtual/jre in DEPEND 16:17 <@Betelgeuse> ssuominen: it confuses the vm switching code 16:17 <@Betelgeuse> ssuominen: the latter is what you should do 16:17 < ssuominen> ah, right the DEPEND=$RDEPEND... 16:17 <@Betelgeuse> you are telling it that building with a jre is fine 16:17 < ssuominen> propably want to use COMMON_DEPEND and separate the RDEPEND with it 16:17 < ssuominen> right? 16:18 <@Betelgeuse> ssuominen: or RDEPEND="${RDEPEND} virtual/jre-1.5" after declaring DEPEND 16:18 < ssuominen> nod 16:18 <@Betelgeuse> We should add a eqawarn to our eclasses if there's both virtual/jdk and virtual/jdk atoms 16:18 <@Betelgeuse> jdk && jre