[paludis-dev] [Paludis] #1342: dev-java/jtidy - build failed (Gentoo)

Paludis trac-paludis at exherbo.org
Sun Mar 6 16:19:45 UTC 2016


#1342: dev-java/jtidy - build failed (Gentoo)
-----------------------------+--------------------
    Reporter:  kapsh         |      Owner:
        Type:  defect        |     Status:  closed
    Priority:  Sometime      |  Milestone:
   Component:  clients/cave  |    Version:  2.4.0
  Resolution:  upstream      |   Keywords:
  Blocked By:                |   Blocking:
Distribution:  Gentoo        |
-----------------------------+--------------------
Changes (by dleverton):

 * status:  new => closed
 * resolution:   => upstream


Comment:

 Replying to [ticket:1342 kapsh]:
 > Portage (I've checked with `ebuild jtidy-1.0.ebuild install`) and even
 direct command `ant jar` can complete this build. I'm not sure this is
 paludis defect or bad written ebuild so I will be glad for the
 explanations.

 That would be because Paludis sets the locale for the build environment to
 C, whereas Portage lets it leak in from the calling environment.  It would
 fail in Portage (or when building manually) for anyone who happened not to
 be using a UTF-8 locale.

 Replying to [comment:1 kapsh]:
 > Paludis also regenerated packaged build.xml (different files attached)
 but `ant jar` eat this too.

 Paludis itself doesn't do anything like that.  The Gentoo Java eclasses
 do, but I don't know the details of how that works, or why it might behave
 differently under Paludis than Portage.

 > **PS** Workaround with adding `encoding="utf-8"` to `javac` target in
 build.xml (like in patch below) worked for me. Is it reasonably to
 generate this attribute too?

 Not sure what you mean by "generate"; if it's related to Gentoo's
 build.xml rewriting then I'd suggest asking the Java team.  But that does
 seem like a sensible fix - if the sources are in UTF-8 then the build
 system should say so, rather than hoping that the user has a compatible
 locale setting.

 > **PPS** And tell me please - should I post this to Gentoo bugzilla too?

 Yes - actually I think it would be an upstream bug, but the proper
 procedure would be to send it to Gentoo first.  (''Actually'' I'd call it
 a bug in the Java compiler for using user settings to parse the code in
 the first place, but I doubt Oracle would be cooperative on that point.)

-- 
Ticket URL: <http://paludis.exherbo.org/trac/ticket/1342#comment:2>
Paludis <http://paludis.exherbo.org/>
Paludis, the Other Package Mangler


More information about the paludis-dev mailing list