Bug 266261

Summary: multimedia/x265: fails to build with default options
Product: Ports & Packages Reporter: f451 <f451>
Component: Individual Port(s)Assignee: Fernando Apesteguía <fernape>
Status: Closed FIXED    
Severity: Affects Some People CC: f451, fernape, freebsd-2024, void
Priority: --- Keywords: needs-patch
Version: LatestFlags: fernape: maintainer-feedback? (freebsd-2024)
Hardware: arm64   
OS: Any   
Attachments:
Description Flags
poudriere testport failure log none

Description f451 2022-09-06 22:20:48 UTC
Created attachment 236404 [details]
poudriere testport failure log

Hello,

multimedia/x265 fails to build with default options which are

     ASM=on: Use optimized assembly code
     DEBUG=off: Enable debugging
     OPTIMIZED_FLAGS=on: Enable O3 optimization
     SVTHEVC=off: HEVC encoding via SVT-HEVC
     VMAF=off: Per frame and aggregate VMAF score
     Options available for the multi PixelWidth: you have to choose at least one of them
     HI8P=on: Enable 8-bit pixel-width (you, probably, want this)
     HI10P=on: Enable 10-bit pixels (may break on i386)
     HI12P=on: Enable 12-bit pixels (may break on i386)

Temporary workaround is to disable the HI10P and HI12P options. 
I didn't test whether it would build with just HI10P off

Build failure message was

[00:03:23] FAILED: common/CMakeFiles/common.dir/aarch64/asm-primitives.cpp.o
 
full poudriere testport log is zipped/attached
Comment 1 Fernando Apesteguía freebsd_committer freebsd_triage 2022-09-16 11:41:48 UTC
I can build multimedia/x265 properly in 12.3, 13.1 and 14-current in i386 and amd64 with default options.

Also, there are no pkg-fallout messages from the builder. This might seem a local problem in arm64.

Could you test which one of the options makes this fail? That way we could mark as BROKEN for that combination in arm64. Unfortunately I don't have an arm64 board to try this.

Cheers.
Comment 2 void 2022-09-16 15:43:44 UTC
OK, I should have some results for you in the next 24 hrs
Comment 3 f451 2022-09-18 03:15:59 UTC
(In reply to Fernando Apesteguía from comment #1)

With ASM on the build will fail due to building gcc-11 which gets killed by OOM. With ASM off the build succeeds with all the HI* options enabled. This is with a fresh poudriere jail and ports tree.

This is different to the issue initially reported, which I've not been able to reproduce. Close fault?
Comment 4 Fernando Apesteguía freebsd_committer freebsd_triage 2022-09-18 15:54:48 UTC
(In reply to f451 from comment #3)
Yes, being killed by the OOM killer is more and more common lately, but as you pointed out that is a different problem.

Since we can not reproduce this (neither me, you or the cluster), I think we should close this as "Unable to reproduce".
Comment 5 Mikhail T. 2022-09-18 21:39:00 UTC
(In reply to f451 from comment #3)
> gcc-11 which gets killed by OOM

Sounds like a problem with lang/gcc11 -- or the machine doing the building... Not enough RAM or swap?

Official build-cluster is not reporting problems -- but offers fresh builds for ARM...
Comment 6 Fernando Apesteguía freebsd_committer freebsd_triage 2022-09-19 07:32:34 UTC
^Triage: PRs should have an assignee (real person) when closing. Assign to myself (triage team member)