Package Details: sentencepiece 0.2.1-2

Git Clone URL: https://aur.archlinux.org/sentencepiece.git (read-only, click to copy)
Package Base: sentencepiece
Description: Unsupervised text tokenizer for Neural Network-based text generation
Upstream URL: https://github.com/google/sentencepiece
Licenses: Apache-2.0
Provides: libsentencepiece.so, libsentencepiece_train.so
Submitter: Henry-ZHR
Maintainer: Henry-ZHR
Last Packager: Henry-ZHR
Votes: 9
Popularity: 0.065906
First Submitted: 2024-03-10 06:42 (UTC)
Last Updated: 2026-02-15 06:48 (UTC)

Pinned Comments

Henry-ZHR commented on 2024-03-26 14:38 (UTC)

For everyone:

You may need to rebuild the package when its dependencies (currently abseil-cpp, protobuf and gperftools) are upgraded.

You can install rebuild-detector so that you will be warned when you need to do so.

Latest Comments

1 2 3 Next › Last »

Henry-ZHR commented on 2026-02-17 09:24 (UTC)

@MLXQt Arch Linux has split gcc-libs into libgcc, libstdc++ and more

You should

  1. Select an up-to-date mirror and then pacman -Syu (if official Arch Linux)
  2. Or wait for your distribution to update gcc-libs

Actually, from 0.2.1-1 to 0.2.1-2, the package's content doesn't change, so you can ignore the update for now

MLXQt commented on 2026-02-17 09:16 (UTC) (edited on 2026-02-17 09:20 (UTC) by MLXQt)

A couple of days ago, I had to update these two packages: python-sentencepiece from 0.2.1-1 to 0.2.1-2 and sentencencepiece as above. The update failed because it couldn’t find the libgcc and libstdc++ dependencies. The problem was that these two dependencies were already included in gcc, a package in the official repositories, so the update failed. A couple of days later, the two sentencepiece packages were updated. The problem was that this update also produced errors:

Conflicting files:

  • libstdc++-snapshot: /usr/lib/libstdc++.so already exists in the filesystem (owned by gcc-libs)
  • libstdc++-snapshot: /usr/lib/libstdc++.so.6 already exists in the filesystem (owned by gcc-libs)
  • libstdc++-snapshot: /usr/share/locale/de/LC_MESSAGES/libstdc++.mo already exists in the filesystem (owned by gcc-libs)
  • libstdc++-snapshot: /usr/share/locale/fr/LC_MESSAGES/libstdc++.mo already exists in the filesystem (owned by gcc-libs)
  • libgcc-snapshot: /usr/lib/libgcc_s.so already exists in the filesystem (owned by gcc-libs)
  • libgcc-snapshot: /usr/lib/libgcc_s.so.1 already exists in the filesystem (owned by gcc-libs)

Again, the dependencies are already included in gcc and gcc-libs and the update fails. Any solutions?

lahwaacz commented on 2025-11-20 18:59 (UTC)

You can replace the _tag with pkgver:

diff --git a/PKGBUILD b/PKGBUILD
index 37d3e9d..ebfe239 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -12,22 +12,17 @@ makedepends=('git' 'cmake'
              'abseil-cpp' 'gperftools' 'protobuf'
              'python' 'python-build' 'python-setuptools' 'python-wheel' 'python-installer')
 checkdepends=('python-pytest')
-_tag='31646a467d2051eb904e0b45de3a73e91fe1c1e3' # git rev-parse "v${pkgver}"
 source=(
-  "${_name}::git+${url}.git#tag=${_tag}"
+  "${_name}::git+${url}.git#tag=v$pkgver"
   'fix-parse-command-line-flags-test.patch'
   'dont-include-data-files-in-python-pkg.patch'
 )
 sha512sums=(
-  'SKIP'
+  'f9bfaf0183da4f1aa27f169b28fb0db95cdc1ff7e3e12f04cc64eae563c0ab5ed6a30e820659736509ca420363473ece1af27657975095080e772d841e5f3475'
   '7329caa5e824b873e931fd10134e07b8c6bbb0612e5355be0ddf8478677d2cd4e2d725485c51173142fcc35b6c12b869aecca6949291b869f537f4d6c048cc55'
   '86443111c23d1d0f7db27673d8bb34e1a5d7e49e301b0e0978bb32b29cde51a924ff95bd079923f993c8e1180400b4268c39fa500102c99925daaaffeefe90ff'
 )

-pkgver() {
-  git -C "${_name}" describe --tags | sed 's/^v//'
-}
-
 prepare() {
   cd "${_name}"

davidovitch commented on 2025-09-27 08:33 (UTC)

@Henry-ZHR apologies for the noise. I was using aurutils for building and it seems there's some issue caused there that I don't understand. If I use plain makepkg all is good.

Henry-ZHR commented on 2025-09-27 08:21 (UTC)

@davidovitch Log please.

Anyway, both tags should refer to the same thing.

davidovitch commented on 2025-09-27 08:16 (UTC) (edited on 2025-09-27 08:17 (UTC) by davidovitch)

Seems like the git hash _tag='31646a467d2051eb904e0b45de3a73e91fe1c1e3 is no longer valid? Build passes tests when using the sources associated with the v0.2.1 tag.

rubin55 commented on 2025-08-30 23:17 (UTC)

just fyi, when you git clone python-sentencepiece (instead of sentencepiece), you seem to get an old copy of this repository, from last february, version 0.2.0.

Henry-ZHR commented on 2025-05-05 13:54 (UTC)

@unit73e Should be fixed now. Thank you for your patch.

unit73e commented on 2025-05-04 16:18 (UTC)

@Henry-ZHR to make it easy: https://github.com/unit73e/sentencepiece/commit/80e07aeb876374530056cb629a2296ee911b1448

1 2 3 Next › Last »