[Bug 244669] devel/llvm10: clang-move should be EXTRAS, not CLANG in pkg-plist; breaks install
bugzilla-noreply at freebsd.org
bugzilla-noreply at freebsd.org
Sun Mar 8 08:07:56 UTC 2020
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=244669
Bug ID: 244669
Summary: devel/llvm10: clang-move should be EXTRAS, not CLANG
in pkg-plist; breaks install
Product: Ports & Packages
Version: Latest
Hardware: Any
OS: Any
Status: New
Severity: Affects Some People
Priority: ---
Component: Individual Port(s)
Assignee: brooks at FreeBSD.org
Reporter: greenreaper at hotmail.com
Assignee: brooks at FreeBSD.org
Flags: maintainer-feedback?(brooks at FreeBSD.org)
Created attachment 212248
--> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=212248&action=edit
Assign bin/clang-move to EXTRAS (not CLANG)
When installing llvm10-10.0.0.r3 after a build with portinstall, using options
CLANG GOLD LLD LLD_LINK OPENMP PYCLANG (not EXTRAS) and BE_NATIVE (X86),
installation fails to register because clang-move can't be find by pkg-static:
```
===> Registering installation for llvm10-10.0.0.r3
pkg-static: Unable to access file
/tmp/usr/ports/devel/llvm10/work/stageusr/local/llvm10/bin/clang-move:No such
file or directory
*** Error code 74
```
(Incidentally this error message appears to be missing a / between
'/tmp/usr/ports/devel/llvm10/work/stage' and 'usr/local/llvm/bin/clang-move')
The reason for this is that clang-move is in clang-tools-extra:
https://github.com/llvm-mirror/clang-tools-extra
...but the binary is marked as being CLANG, not EXTRAS, in pkg-plist:
%%CLANG%%llvm%%LLVM_SUFFIX%%/bin/clang-move
Changing this to:
%%EXTRAS%%llvm%%LLVM_SUFFIX%%/bin/clang-move
resolves the problem.
lib/libclangMove.a is not affected as it is already EXTRAS:
%%EXTRAS%%llvm%%LLVM_SUFFIX%%/lib/libclangMove.a
This change was introduced in revision 525893:
https://svnweb.freebsd.org/ports/head/devel/llvm10/pkg-plist?r1=525893&r2=525892&pathrev=525893
...submitted via review D23583 (Fix build for LLVM 10.0.0rc1 extras):
https://reviews.freebsd.org/D23583
The test plan in the review appears to use the default options (all components)
which will not catch a failure to segregate binaries or libraries, just the
failure to build one when everything else is built. Testing with/without EXTRAS
would have been a good idea, since the goal was specifically to add EXTRAS
back.
More generally, this isn't the first time a build without EXTRAS has failed on
install, and it might be good to have an automated test of some kind that
catches this class of bug (perhaps with other combinations of install flags).
--
You are receiving this mail because:
You are the assignee for the bug.
More information about the freebsd-ports-bugs
mailing list