Bug 210991
| Summary: | astro/gpsbabel: Fix build with libc++ 3.8.0 | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Oliver Heesakkers <dev2> | ||||||
| Component: | Individual Port(s) | Assignee: | Thomas Zander <riggs> | ||||||
| Status: | Closed FIXED | ||||||||
| Severity: | Affects Some People | CC: | riggs | ||||||
| Priority: | --- | Keywords: | patch | ||||||
| Version: | Latest | Flags: | riggs:
merge-quarterly+
|
||||||
| Hardware: | Any | ||||||||
| OS: | Any | ||||||||
| See Also: | https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=208158 | ||||||||
| Attachments: |
|
||||||||
Thanks Oliver. Can you please: * Confirm this passes QA (portlint, poudriere) * Set maintainer-approval on attachments for ports you are MAINTAINER of Yes, Poudriere builds of i386 and amd64 for 1100120 were OK, I'll run more later today. portlint says "looks fine" Created attachment 172371 [details]
New patch which eliminates a warning in most build environment
Updated patch. By adding (int) before the value to print, exif.cc now builds without warning on all tested platforms except 9 i386, where a warning about comparing signed and unsigned persists
Other than that port build succesfully on:
JAILNAME VERSION ARCH METHOD TIMESTAMP
10amd64 10.3-RELEASE-p5 amd64 ftp 2016-07-11 11:01:40
10i386 10.3-RELEASE-p5 i386 ftp 2016-07-11 11:03:37
11amd64 11.0-BETA1 amd64 ftp 2016-07-10 22:50:37
11i386 11.0-BETA1 i386 ftp 2016-07-10 22:26:48
9amd64 9.3-RELEASE-p44 amd64 ftp 2016-07-11 11:08:38
9i386 9.3-RELEASE-p44 i386 ftp 2016-07-11 11:13:55
portlint is OK
Thank you Oliver Don't forget to obsolete the previous patch :) Click Attachment -> Details -> Edit Details -> [X] Obsolete Not sure what this PR is waiting for now. I've kind of had it with all the pkg fallout mails. A commit references this bug: Author: riggs Date: Sun Aug 28 07:42:58 UTC 2016 New revision: 421009 URL: https://svnweb.freebsd.org/changeset/ports/421009 Log: Fix build with libc++ 3.8.0 PR: 210991 Submitted by: dev2@heesakkers.info (maintainer) MFH: 2016Q3 (ports-secteam build fix blanket) Changes: head/astro/gpsbabel/files/patch-exif.cc A commit references this bug: Author: riggs Date: Sun Aug 28 07:45:53 UTC 2016 New revision: 421010 URL: https://svnweb.freebsd.org/changeset/ports/421010 Log: MFH: r421009 Fix build with libc++ 3.8.0 PR: 210991 Submitted by: dev2@heesakkers.info (maintainer) Approved by: ports-secteam (build fix blanket) Changes: _U branches/2016Q3/ branches/2016Q3/astro/gpsbabel/files/patch-exif.cc |
Created attachment 172361 [details] patch to patch exif.cc Ever since libc++ 3.8.0 was introduced, I've had pkg fallout bugging me about this. See #208158 Initially I wanted to combine the update to 1.5.3 with a fix for this problem, but to my surprise the update was already done through phabricator by the time I got to it (I know I'm late to update, but not even a courtesy e-mail?). The original complaint from pkg fallout was "call to 'abs' is ambiguous" and only leads to a failed build under i386. Replacing abs() with labs(), as is done upstream, would fix the build, but warnings would still be displayed. Since the problem is presumably caused by QT's toTime_t() I then figured it would be appropriate to replace abs() with QT's qAbs(). The build now completes without error or warning on i386 and completes without error but with a warning on amd64 ("format specifies type 'int' but the argument). Since this warning is only on printing a warning I figured I'd get patch in as is and try to address this warning later when I get the chance. Tested against i386 and amd64 1100120.