Bug 250595
| Summary: | devel/hyperscan: update 5.2.1 -> 5.3.2 | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Kurt Jaeger <pi> | ||||||||
| Component: | Individual Port(s) | Assignee: | Vsevolod Stakhov <vsevolod> | ||||||||
| Status: | Closed Overcome By Events | ||||||||||
| Severity: | Affects Only Me | CC: | diizzy, pi | ||||||||
| Priority: | --- | Flags: | bugzilla:
maintainer-feedback?
(vsevolod) |
||||||||
| Version: | Latest | ||||||||||
| Hardware: | Any | ||||||||||
| OS: | Any | ||||||||||
| Attachments: |
|
||||||||||
|
Description
Kurt Jaeger
2020-10-25 16:10:52 UTC
Is it possible to bump boost to 1.74.0 to match upcoming version in tree? https://reviews.freebsd.org/D24485 While I understand that maintaining external patches isn't ideal it might be of interest to possibly import (aarch64 support) https://github.com/MarvellEmbeddedProcessors/hyperscan/commits/master Unfortunately upstream doesn't seem to be too keen about the idea :( https://github.com/intel/hyperscan/issues/197 https://www.vectorcamp.gr/2020/05/09/we-are-thrilled-to-announce-our-first-customer/ fwiw, Alpine Linux uses their fork now (In reply to daniel.engberg.lists from comment #1) Using the patch with Patching file src/util#PATCH_SITES= https://github.com/MarvellEmbeddedProcessors/${GH_PROJECT}/commit/ #PATCHFILES= 8a786129edb297285820ee1963de2472e38ea820.patch:-p1 fails with this: Patching file src/util/simd_utils.h using Plan A... Hunk #1 succeeded at 33. Hunk #2 failed at 73. Hunk #3 failed at 81. 2 out of 3 hunks failed--saving rejects to src/util/simd_utils.h.rej (In reply to daniel.engberg.lists from comment #2) Ah, now I get it. VectorCamp has a fork 5.3.2, where the patch is already included. Indeed, here's how Alpine does it (in case it helps) https://git.alpinelinux.org/aports/tree/community/hyperscan/APKBUILD Created attachment 221141 [details]
patch
testbuild on 13 is ok
and changing upstream from intel to VectorCamp as suggested (In reply to daniel.engberg.lists from comment #5) 5.3.2 builds, just not on aarch64 8-( https://people.freebsd.org/~pi/logs/hyperscan-5.3.2.log Created attachment 221362 [details] Build log (failed) on arm64 https://github.com/VectorCamp/hyperscan/blob/v5.3.2/CMakeLists.txt#L247 https://github.com/VectorCamp/hyperscan/blob/v5.3.2/CMakeLists.txt#L251 Needs to be removed or at least changed as native doesn't work on aarch64 but it still fails to build :/ I don't have aarch64 with FreeBSD to test :( I will try to ask br@ as he definitely has one (or more). (In reply to Vsevolod Stakhov from comment #10) With the suggested modification for CMakeLists.txt, it gets over the initial problem, but fails at some other part: https://people.freebsd.org/~pi/logs/hyperscan-arm-v2.txt I guess we could create a new port as they've decided to change the name https://github.com/VectorCamp/vectorscan/releases/tag/vectorscan-v5.4.0 fwiw, compiles using GCC on 13-CURRENT (aarch64) Vsevolod Could you consider adoptiong vectorscan instead so we could also bring support to arm64 and eventually Power too? Updated to 5.4.0 in commit a52662a72f5541c0d871f17c46832e7726657a87 |