Summary: | net-dns/bind-tools-9.4.3_p2: fix cross-compilation | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Bertrand Jacquin <bertrand> |
Component: | Current packages | Assignee: | BIND Maintainers (DISABLED) <bind+disabled> |
Status: | RESOLVED FIXED | ||
Severity: | normal | ||
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
bind-tools-BJA-epoll-AC_TRY_RUN-cross.diff
Use HOSTCC to build gen net-dns/bind-tools-9.4.3_p2.ebuild |
Description
Bertrand Jacquin
2009-04-17 19:28:54 UTC
Created attachment 188713 [details, diff]
bind-tools-BJA-epoll-AC_TRY_RUN-cross.diff
Created attachment 188715 [details, diff]
Use HOSTCC to build gen
Created attachment 188717 [details, diff]
net-dns/bind-tools-9.4.3_p2.ebuild
patch have been reported upstream : https://lists.isc.org/pipermail/bind-workers/2009-April/002780.html Patch for configure.in (not present here as configure is used as is, but send upsteam) have been integrated upstream : https://lists.isc.org/pipermail/bind-workers/2009-April/002784.html Hi, From what i can see only the second patch is required for building crosscompiled bind 9.6.1, could you please confirm? :] Exact, first one have been included upstream : https://lists.isc.org/pipermail/bind-workers/2009-April/002784.html Reading the build system files I tend to say using CC_FOR_BUILD is better than using HOST_CC. Do you agree? Btw. can you please test more recent versions too? I also think that net-dns/bind is also affected by this bug. I need to do some tests with more recent release. I don't need bind package but you're right, bind also apply in this bug report. Also, CC_FOR_BUILD is used to build destination binary, not strictly host binary like "gen" which is used and executed in building process Bertrand Its just that I didn't saw HOST_CC there so I wonder how/where it will be defined? this affects bind-tools-9.7.1. attached patches allow compilation with some rewriting to account for altered files For the record, CC_FOR_BUILD is the CC for ${CBUILD} handling, which is what should be used for tools then used during build. The AC_TRY_RUN can be avoided, at least as of 9.8.0-P1, by adding this to econf call $(use_enable kernel_linux epoll) which avoids running it in "auto" mode. And BUILD_CC should be exported as $(tc-getCC_FOR_BUILD) Okay I fixed this one at least partially. USE=ssl fails because of another AC_TRY_RUN call, but I'll see to write some documentation about AC_TRY_RUN, so that I can tell upstream how to fix this. |