Summary: | security/softether: Missing runtime library libiconv | ||||||||
---|---|---|---|---|---|---|---|---|---|
Product: | Ports & Packages | Reporter: | Franco Fichtner <franco> | ||||||
Component: | Individual Port(s) | Assignee: | Kurt Jaeger <pi> | ||||||
Status: | Closed FIXED | ||||||||
Severity: | Affects Only Me | CC: | koobs, marino, net, pi | ||||||
Priority: | --- | Keywords: | easy, patch | ||||||
Version: | Latest | Flags: | pi:
maintainer-feedback+
pi: merge-quarterly+ |
||||||
Hardware: | Any | ||||||||
OS: | Any | ||||||||
Attachments: |
|
Description
Franco Fichtner
2015-10-31 23:22:49 UTC
This change (if verified) has blanket/implicit approval. I was investigating another issue with this port (it's JOBS UNSAFE always, not just with openSSL) and notices that USES=iconv is *NOT* used. This looked home-grown to me. I suspect using BUILD_DEPENDS is a mistake that should have been caught when port was introduced last month. I'm CC'ing pi@ who brought it in. To summarize: 1) I think USES=iconv should be used 2) This needs MAKE_JOBS_UNSAFE=yes unconditional (It tries to build object files before /tmp/obj/ directory is created. If 1) is wrong, the maintainer should explain specifically why libiconv handling differs from directions in Mk/Uses/iconv.mk I just had a look at the port: USES=iconv is in the Makefile, but when I build it without the BUILD_DEP on iconv, it fails to build on 11 and 10. What do you mean with "USES=iconv is not used" ? I didn't see it -- I didn't look because I assumed it wasn't there since BUILD_DEPENDS was set. If it fails on 10 and 11, then I assume some of the configuration args defined in Uses/iconv.mk need to be used. I suspect this > F10 is a hack that works around using iconv.mk as it was intended. I just tried it in poudriere with USES=iconv:build, but if I do not add the BUILD_DEP, it fails. I don't think you got what I said. Take a look at iconv.mk. There are many variables defined in there that are not used by default. If the port needs that information, you are supposed to add it to CONFIGURE_ARGS, LDFLAGS, etc. If the build is failing, then some configuration info is missing. the ":build" thing is the wrong approach. This PR just confirms that it needs to be set in LIBS if used. another thing you try: USES= iconv:translit or USES= iconv:wchar_t I just hit an unsafe jobs issue as well. Doesn't always happen, but often enough to cause annoyance. Sorry for the late reply but the real life needed some more attention then anything computing related. I'll look into this issue, also the other bug reported related to the OpenSSL building problems. When originally porting, testing was done on 9.3 / 10.0 / 10.1 (both amd64 and i386) and there hasn't been issues. Created attachment 162698 [details] Update for iconv / bug 203688 Created attachment 162699 [details] svn diff to address iconv issue and bug 203688 Sorry - incorrectly used "File" instead of "Comment" The attached svn diff should address the iconv problems, as well as https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=203688 Tested unter 9.3 / 10.1 - make pkg pulls in converters/libiconv as dependency. Hope that fixes the problems. so iconv:wchar_t worked to build on all platforms? I assume pi@ is still taking care of these softether PRs. You need to submit a new version. Softether is jobs *unsafe*, period. MAKE_JOBS_UNSAFE=yes needs to be set unconditionally. We also need a revision bump. :) testing@work I'll add the PORTREVISION and the JOBS_UNSAFE, no need for additional patches. FYI it's *move" JOBS_UNSAFE, not "add" A commit references this bug: Author: pi Date: Mon Nov 2 12:00:00 UTC 2015 New revision: 400650 URL: https://svnweb.freebsd.org/changeset/ports/400650 Log: security/softether: fixes - iconv build - build with No-SSLv3 - MAKE_JOBS_UNSAFE=yes unconditionally PR: 203688, 204184 Submitted by: net@arrishq.net (maintainer) Changes: head/security/softether/Makefile head/security/softether/files/patch-src__Mayaqua__Network.c Committed, thanks for the bug reports and the patch! @Kurt MFH? I'll MFH. I didn't notice this before: "iconv:translit iconv:wchar_t" It's one or the other, not both. Well, the last one wins, as we learned from the 'USES should be sorted, but has side effects beyond that' thread. So... A commit references this bug: Author: pi Date: Mon Nov 2 14:11:09 UTC 2015 New revision: 400655 URL: https://svnweb.freebsd.org/changeset/ports/400655 Log: security/softether: fix USES PR: 204184 Submitted by: marino Approved by: maintainer (implicit) Changes: head/security/softether/Makefile A commit references this bug: Author: pi Date: Mon Nov 2 19:44:44 UTC 2015 New revision: 400670 URL: https://svnweb.freebsd.org/changeset/ports/400670 Log: MFH: r400650 - iconv build - build with No-SSLv3 - MAKE_JOBS_UNSAFE=yes unconditionally PR: 204184 Approved by: ports-secteam (feld) Changes: branches/2015Q4/security/softether/Makefile branches/2015Q4/security/softether/files/patch-src__Mayaqua__Network.c If I'd believe in vampires, I'd apply silver bullets now 8-} |