Quote:
and https://roy.marples.name/git/dhcpcd EDIT: Roy is on holiday, so we'll wait a bit, at least for a reply |
From Roy:
Code:
Looks like we need to add a define for __NR_getrandom in privsep-linux.c |
The problem with the kdepim mailcommon package and cmake is solved with the new version of cmake-3.24.1
|
Quote:
Code:
[PATCH] privsep: Allow getrandom sysctl for newer glibc EDIT: just tested and, of course, it works + Code:
privsep: Allow newfstatat syscall as well |
ok, included in the SFS build tree.
|
apparently there's another patch concerning glibc-2.36 : https://github.com/NetworkConfigurat...733585d74c4ff5
with this patch : https://github.com/NetworkConfigurat...d74c4ff5.patch updated SFS source build tree. created a git version of dhcpcd which works: https://github.com/nobodino/slackwar...cd-git-version |
Quote:
|
you're right.
Thanks. |
Regression test up to "Fri Oct 14 01:39:37 UTC 2022" (x86_64 version).
Just 2 remaining packages don't build correctly: - f2fs need a patch to build with lz4-1.9.4 (see: https://www.linuxquestions.org/quest...ls-4175715792/) Code:
diff -aurN f2fs-tools-1.15.0/fsck/compress.c f2fs-tools-1.15.0-mod/fsck/compress.c here' what's left: Code:
--- list-slackware 2022-10-16 09:17:41.244788733 +0200 |
Regression test up to "Thu Oct 27 02:30:15 UTC 2022" (x86_64 version).
Not too bad, only one remaining package refusing to build. Code:
--- list-slackware 2022-10-29 14:42:21.062632918 +0200 |
Regression test up to "Mon Oct 31 23:31:36 UTC 2022" (x86_64 version).
Except glibc and s-nail packages which had problems with new make package, nothing new under the sun. |
Nethack no more build from gcc 12 (not detected because the package was build also with errors).
Fixing config.h: --- nethack.SlackBuild.orig 2022-09-15 20:23:58.052165083 +0200 +++ nethack.SlackBuild 2022-12-12 17:04:44.146540910 +0100 @@ -107,6 +107,9 @@ sed -e "/^MANDIR\s*=/s|/usr/man/man6|$PKG/usr/man/man6|" -i sys/unix/Makefile.doc +# Fixing gcc 12 compatibility: +sed -i 's/#define UNIX/#include <stdlib.h>\n#include <stdio.h>\n&/' include/config.h + # Create a few directories: mkdir -p $PKG/usr/man/man6 mkdir -p $PKG/usr/doc/nethack-$VERSION @@ -116,7 +119,7 @@ ( cd sys/unix ; sh setup.sh hints/linux ) # Build and install: -make -j1 install manpages PREFIX=$PKG +make -j1 install manpages PREFIX=$PKG || exit 1 # Fix some paths: sed -e "s|HACKDIR=$PKG/|HACKDIR=/|" \ |
Quote:
https://github.com/NetHack/NetHack/c...7442ed7ff2fe4e |
I do think libepoxy should be done, and it shouldn' be difficult.
gelf.h is not required for <=Xorg-22 nor <=mesa-22, it must be an option of something optional (GL and firefox work without it, as for the rest IDK) > 6) Also had to recompile glib2 before gtk-pixbuf2 would compile. perhaps that is because you inadvertently changed the order? or maybe slack-build is not an "error free build" ? ======================== if you recompile Xorg after the build (or rather proceed forward then recompile it) YOU WILL need to recompile xf86-input-* xf86-video-*, since they are coded to not load unless built after Xorg |
@xlfs-0.2: I don't understand a single word of what you're talking about, sorry.
I looked at your profile, you're trying to build XLFS, that's fine. I'll look at your project during christmas holydays, but clearly I don't understand your message! |
Regression test up to "Sun Dec 25 23:21:25 UTC 2022" (x86_64 version).
Nothing special, except the qt5-webkit package (problem solved in another thread) and always the same pangomm. |
@nobodino
When do you recompile the entire stack ? Do you plan it, or after every updates ? |
For now, Slackware has become very solid according to the possibility to be rebuilt from scratch, it doesn't need to be recompiled with every bit of software changed.
I only do it on the week-end and if I have enough time to do it. Now I recompile it every time: - one of the glibc / binutils or gcc change: a special attention to glibc with deprecated features (twice a year for glibc and binutils) - one of the automake / autoconf / make /texinfo change: those gnu packages are not always safe as upgrade. - rust change because of its impact on firefox / thunderbird and seamonkey. - python3 change, if python3-3.10.x comes back or python3-3.11.x - if while building the "tools" something odd occures, I'm sure I've to recompile the all set. Otherwise, if I waited a too long time, around a month, I give it a try. |
Regression test up to "Sun Jan 15 01:02:36 UTC 2023" (x86_64 version) with binutils-2.40.
Except: - qt5-webkit package (problem solved in another thread) - always the same pangomm - gdb (problem solved in another thread) Nothing else. |
Regression test up to "Fri Feb 3 20:04:33 UTC 2023" (x86_64 version).
Except: - qt5-webkit (same problem) - pangomm (same problem) |
With .so version bump for the latest libvpx upgrade (1.12: libvpx.so.7 => 1.13: libvpx.so.8), some packages should be rebuilt:
- ffmpeg - gst-plugins-good - qt5 - xine-lib Thanks! |
Quote:
https://www.linuxquestions.org/quest...nt-4175698890/ |
Quote:
|
If vulkan-sdk is compiled after qt5 (as needed if you want to have qt5-based tools), compiling VulkanTools will fail if gcc>=12 due to uninitialized variables.
Simple "patch" to solve the problem: --- a/source/x/vulkan-sdk/vulkan-sdk.SlackBuild +++ b/source/x/vulkan-sdk/vulkan-sdk.SlackBuild @@ -383,6 +388,9 @@ tar xvf $CWD/VulkanTools-sdk-$VULKANTOOLS_VERSION.tar.?z || exit 1 cd $TMP/VulkanTools-sdk-$VULKANTOOLS_VERSION +# Fix warning-error on gcc 12 about uninitialized variables: +sed -i 's/Value v;/Value v(0);/' submodules/jsoncpp/src/lib_json/json_reader.cpp + chown -R root:root . find . \ \( -perm 777 -o -perm 775 -o -perm 711 -o -perm 555 -o -perm 511 \) \ |
1 Attachment(s)
From sqlite 3.41, double-quotes are not more accepted by default as string literal.
Package ibus-libpinyin uses double-quotes for string literal, so if compile after sqlite3 (as in Slackware from Scratch) it will fail to build english.db. Already fixed in github repository: https://github.com/libpinyin/ibus-li...88942af438ea98 |
cups-filters does not compile against newer qpdf that uses std::string_view (available only from c++17).
There are 2 possible solutions: 1) Replace in the build script: Code:
CXXFLAGS="$SLKCFLAGS -std=c++11 -fpermissive" Code:
CXXFLAGS="$SLKCFLAGS -std=c++17 -fpermissive" Code:
--- cups-filters-1.28.17/Makefile.am.orig |
Quote:
Code:
ap/FTBFSlog |
Regression test up to "Fri Mar 17 20:12:03 UTC 2023" (x86_64 version), which is on "Happy Saint Paddy's Day!"
Very few regression, and one is already closed: - the green one - the blue one has 2 solutions, one from me, another from marav. Code:
--- list-slackware 2023-03-19 11:15:38.008100233 +0100 |
Regression building mesa after upgrading rust-bindgen to 0.65.
Solution in thread https://www.linuxquestions.org/quest...-a-4175724025/ |
Regression test up to "Fri Apr 28 18:25:13 UTC 2023" (x86_64 version) with gcc-13.1.
Not so bad but, here are some non building packgages: - nodejs: problem solved by marav - ncftp: problem solved by marav - mariadb: problem solved by marav - vulkan-sdk: no complete solution - qt5webkit: problem solved - qt-5.19.6: problem solved by marav - mozilla-thunderbird: problem solved by marav - python2-module-scollection: need to upgrade M2crypto package to 0.38.0 Specific to SFS: - llvm: doesn't build on SFS with gcc-13.1 for the 1st pass |
Quote:
Honestly, I don't know. Mostly others don't seem to compile DirectXShaderCompiler or Vulkan-Profiles and gcc 13 has only just arrived in the testing/ branch of Arch So, wait and see |
fcitx5 does not more compile with fmt 10.0.0 (the latter upgraded on May 10th).
Already fixed in git (but not yet released): https://github.com/fcitx/fcitx5/comm...a17b9b8f6a506b Code:
--- fcitx5-5.0.23/src/modules/dbus/dbusmodule.cpp.orig 2022-11-16 08:56:34.000000000 +0100 |
Regression test up to "Sat Aug 26 21:01:45 UTC 2023" (x86_64 version).
Not tested for a long time. Some problems already solved: - openssh: solved - fcitx5-m17n: solved - MPlayer: a solution by marav, but doesn't solve the root cause which is gettext. Buildable - vulkan: no solution even with newer solution up to 1.3.260.0 Code:
/tmp/VulkanTools-sdk-1.3.204/vkconfig_core/../vkconfig_core/version.h:61:5: note: ‘uint32_t’ is defined in header ‘<cstdint>’; did you forget to ‘#include <cstdint>’? |
Quote:
Code:
--- a/vkconfig_core/version.h |
For me, both mplayer et vulkan FTBFS have been solved now
|
Regression test up to "Fri Sep 1 20:16:14 UTC 2023" (x86_64 version).
Nearly perfect, always the python2-module-collection package but I can trick it. |
All times are GMT -5. The time now is 08:40 AM. |