Bug 221114
| Summary: | devel/bazel fix build on i386 | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Klaus Aehlig <aehlig> | ||||||
| Component: | Individual Port(s) | Assignee: | Richard Gallamore <ultima> | ||||||
| Status: | Closed FIXED | ||||||||
| Severity: | Affects Only Me | CC: | aehlig, amutu, ultima | ||||||
| Priority: | --- | Keywords: | needs-qa | ||||||
| Version: | Latest | Flags: | koobs:
maintainer-feedback+
koobs: merge-quarterly? |
||||||
| Hardware: | Any | ||||||||
| OS: | Any | ||||||||
| Attachments: |
|
||||||||
Example of the log of a failed build: http://beefy10.nyi.freebsd.org/data/110i386-default/446906/logs/bazel-0.5.3.log Thank you Klaus Could you please confirm that this change (and thus the port) passes QA (portlint/poudriere) (In reply to Kubilay Kocak from comment #2) The port passes portlint (the port is a master port, hence the ?= for MAINTAINER is OK and the only occurrence of a plain "file" is in the error message in the BROKEN statement). I don't have a 32-bit system at hand, but I can confirm the following. - The port still build cleanly on am64 and the resulting binary works (not surprising, as nothing changed there). - It also builds cleanly and works if the EXTRA_PATCH is applied unconditionally. What the extra patch does, is switching back to the java-implementation of singlejar that was also used in the previous bazel release everywhere (hence also was successfully used on i386). So I'm pretty confident that this solves the issue, but I didn't do a QA run on i386. Hence I would appreciate, if an i386 QA run would be done. Thank you for clarification Klaus. For what its worth, poudriere does allow testing of i386 (and other arch) on amd64 hosts, and in the long-run is well worth having around and using, particular for confirming correct 'packaging'. (In reply to Klaus Aehlig from comment #3) 103i386: Failed: https://poudriere.ultimasbox.com/data/103i386-test/2017-08-05_21h56m45s/logs/errors/bazel-0.5.3.log 110i386: Failed: https://poudriere.ultimasbox.com/data/110i386-test/2017-08-05_21h56m47s/logs/errors/bazel-0.5.3.log 12i386: Failed: https://poudriere.ultimasbox.com/data/12i386-test/2017-08-05_21h56m49s/logs/errors/bazel-0.5.3.log They are still failing. (In reply to Richard Gallamore from comment #5) Thanks for the report. I'm currently setting up an i386 test environment and will update the patch once I fixed all i386 failures. Created attachment 185169 [details] patch for deve/bazel to fix build on i386 Completed the patch, to also fix the incorrect call to nanosleep. With that patch I could successfully build a working bazel on i386 (FreeBSD 11). The nanosleep patch is also upstreamed, see https://bazel-review.googlesource.com/c/14991 (In reply to Klaus Aehlig from comment #7) Is this this port have slave ports attached to it? Never mind, it does. (In reply to Richard Gallamore from comment #9) yes, it is devel/bazel-clang38 A commit references this bug: Author: ultima Date: Thu Aug 10 18:35:47 UTC 2017 New revision: 447714 URL: https://svnweb.freebsd.org/changeset/ports/447714 Log: * Fix build on i386 * Cleanup Makefile PR: 221114 Submitted by: Klaus Aehlig (maintainer) Reviewed by: matthew (mentor) Approved by: matthew (mentor) MFH: 2017Q3 Differential Revision: https://reviews.freebsd.org/D11937 Changes: head/devel/bazel/Makefile head/devel/bazel/files/extra-i386_src_BUILD Committed, pending MFH authorization. A commit references this bug: Author: ultima Date: Thu Aug 17 21:34:56 UTC 2017 New revision: 448196 URL: https://svnweb.freebsd.org/changeset/ports/448196 Log: MFH: r445987 r446868 r447714 * Update to 0.5.3 * Fix build on i386 * Cleanup Makefile PR: 221114 Submitted by: Klaus Aehlig (maintainer) Reviewed by: matthew (mentor) Approved by: matthew (mentor) Differential Revision: https://reviews.freebsd.org/D11937 Approved by: ports-secteam (feld) Changes: _U branches/2017Q3/ branches/2017Q3/devel/bazel/Makefile branches/2017Q3/devel/bazel/distinfo branches/2017Q3/devel/bazel/files/extra-i386_src_BUILD branches/2017Q3/devel/bazel/files/patch-src_main_java_com_google_devtools_build_lib_bazel_rules_python_python__stub__template.txt branches/2017Q3/devel/bazel/files/patch-src_main_java_com_google_devtools_build_lib_bazel_rules_python_stub__template.txt MFH complete, thanks! |
Created attachment 184864 [details] patch for deve/bazel to fix build on i386 On 32-bit systems, the C-based implementation of singlejar (a tool embedded into bazel) cannot be used, as that implementation assumes (at various places) that it runs on a 64-bit system. Therefore, fall back to the java implementation on those systems.