Summary: | security/p5-Crypt-SSLeay 0.72 fails to build with CC=clang in /etc/make.conf | ||
---|---|---|---|
Product: | Ports & Packages | Reporter: | Trond Endrestøl <Trond.Endrestol> |
Component: | Individual Port(s) | Assignee: | freebsd-perl (Nobody) <perl> |
Status: | Closed Unable to Reproduce | ||
Severity: | Affects Only Me | CC: | adamw, sunpoet |
Priority: | --- | ||
Version: | Latest | ||
Hardware: | Any | ||
OS: | Any |
Description
Trond Endrestøl
2014-07-16 18:04:37 UTC
On stable/10 and head I'm in the (bad?) habit of specifying CPYTYPE?=corei7 in /etc/make.conf, which is acceptable for clang, but not for gcc, at least not for /usr/bin/gcc, as gcc rejects -march=corei7 and -mtune=corei7 when run as part of a test compilation done by the Makefile.PL file. I'm also in the (equally bad?) habit of specifying WITH_GCC=yes and WITH_GNUCXX=yes in /etc/src.conf. Thus my reason to force the selection of clang as the compiler for security/p5-Crypt-SSLeay. Maybe it's simply my fault. I don't know if setting CPUTYPE in /etc/make.conf is discouraged. Maybe CPUTYPE should be moved to /etc/src.conf. over to maintainer I cannot reproduce this problem. I could build this port successfully with CC=clang on my FreeBSD box. I also tried CC=clang36 and it builds without problem. A commit references this bug: Author: adamw Date: Sun Mar 8 19:45:55 UTC 2015 New revision: 380793 URL: https://svnweb.freebsd.org/changeset/ports/380793 Log: Remove obsolete cruft from Makefile that either no longer reflects the codebase, or touches code paths that aren't accessed. PR: 191901 Submitted by: Trond.Endrestol@ximalas.info Changes: head/security/p5-Crypt-SSLeay/Makefile I can also confirm that this port respects CC= just fine, so I'm closing this PR. However, your last line there about the REINPLACE stuff needing rework was quite correct. I yanked the whole thing out (along with other unused variables). Sorry this PR sat for so long, but thank you for submitting it! |