|
Quote:
Might be good to do some tests though. |
Quote:
|
Pat,
Proposed patch for rc.S to create /tmp directories only when needed. I noticed when creating an X-less system. Code:
--- /tmp/rc.S 2021-04-29 17:29:45.498905886 -0500 |
Quote:
yeah, I've had the odd issue over time, been mirroring slackware for 15 years, mirror rsync logs goes back only 10 years, and not once have they been imported till last night, was the only thing that seemed different, I guess a later rsync will clean them up if they aren't supposed to be there or upstream mirror re adjusts their acl's, they dont look nefarious so I'm not bothered, just curious. As for two kernels, did you try again right after, maybe it was in middle of sync? Although I use delete before, many dont. |
libgtop contains a /usr/info/dir that clobbers whatever install-info does.
|
Hi Mr. Volkerding, it seems the wg-quick script (from wireguard-tools) uses resolvconf (a program name) for some of its operation. I don't know much about this to give a good recommendation, but I found resolvconf being provided by openresolv which is available in slackbuilds.org here: https://slackbuilds.org/repository/1...rk/openresolv/
After installing that, wg-quick works as expected (at least for my very simple use case). Also, I think wg and wg-quick needs to be moved to /usr/sbin since they require root permission. But, I don't use wireguard-tools extensively so I'm probably wrong about this. Thanks! |
Don't use a percent character in a filename in the repository
The percent character gets translated *twice* by wget over HTTPS.
Code:
% wget https://mirrors.slackware.com/slackware/slackware64-current/source/l/qt5/patches/qtwebengine-everywhere-src-5.15.2-%25231904652.patch.gzEd |
Request: add PyQtWebEngine since it's been split out of PyQt
While Slackware64 current includes QtWebEngine, PyQtWebEngine is not present. This seems a little odd given that the coverage of the installed components of Qt5 by the installed PyQt5 is otherwise mostly complete. A note on the PyQy website says:
Quote:
Slackware64 current does have PyQtWebKit, but it appears that this is regarded as deprecated in favour of QtWebEngine by Qt. See for example, the second comment in the thread at https://forum.qt.io/topic/55504/anyo...e-of-qt-webkit. It therefore seems prudent to add PyQtWebEngine to Slackware as it will be used by more and more software if QtWebKit/PyQtWebKit is indeed deprecated. There's a SlackBuild script at https://slackbuilds.org/repository/1...PyQtWebEngine/ which I expect could be adapted for Slackware current. There doesn't appear to be much required in the way of changes other than to the PyQtWebEngine version number. |
Hi,
Quote:
I'll try to re-confirm what you've said, especially when it comes to openresolv. Sureley, the commands should go to sbin, as you're correct they require root. I tried wg-quick on Ubuntu sometime ago. It worked no problems there. If openresolv is really required, then I second it's addition. Wiregurad promise is to be ultra easy to set up, and it is indeed with wg-quick, so my opinion is that everything required for wg-quick to work out of the box should be included. -- Best regards, Andrzej Telszewski |
Quote:
I think, probably that : The error in src/utils.c, which was already issued (the ending with the double brackets) : Code:
/* If the FILE itself doesn't exist, return it withoutCode:
15step :: » /usr/local/bin/wget https://mirrors.slackware.com/slackware/slackware64-current/source/l/qt5/patches/qtwebengine-everywhere-src-5.15.2-%25231904652.patch.gzCode:
15step :: ~ » ls qt* |
Quote:
I submitted a bug report on wget https://savannah.gnu.org/bugs/index.php?60494 Pat may want to rename the file if the wget bug does not get fixed soon. Ed |
Quote:
I missed this detail My error was unfortunately confirmed by the fact that the error probably did not exist before ... |
librsvg-2.50.5
https://download.gnome.org/sources/l...-2.50.5.tar.xz |
Bumping request for sof-firmware
Quote:
Summary: sof-firmware is needed in some Lenovo laptops to make speakers and mic work. There are sources (https://github.com/thesofproject/sof) and binaries (https://github.com/thesofproject/sof-bin). I made a slackbuild that repackages the binaries, as compiling the sources is pretty overkill for me and other distros (Arch, Fedora, openSUSE) also repackage the binaries. More info: https://www.linuxquestions.org/quest...ml#post6199638 https://www.linuxquestions.org/quest...5/#post6167184 I'm working on an update to sof-firmware 1.7 but the structure of the source package changed a lot so I'll have to rework my slackbuild (https://github.com/richarson/slackbu...s/sof-firmware). |
at-spi2-core-2.40.1
https://download.gnome.org/sources/a...-2.40.1.tar.xz |
source/installer/ChangeLog.txt:
Code:
Fri Apr 30 17:30:36 UTC 2021 |
In case my mails didn't arrive, please consider updating texlive -> 2021, here it is:
https://ctan.net/systems/texlive/Sou...-source.tar.xz http://slackware.schoepfer.info/slac...-210418.tar.xz http://slackware.schoepfer.info/slac...ve/texlive.tar Some infos to the build changes, -texmf-base is now named YYMMDD, rather than YYYY.YYMMDD -building against system poppler/xpdf isn't possible anymore by upstream -removed unnecessary LD_LIBRARY_PATH and texmf.cnf adjustments before the build, which seems was needed some releases ago -added some fixes by upstream after the release -removed tabu.sty workaround for doxygen, as doxygen is now build without docs -adjusted configure-options to make "make check" work |
sip >= 5.x
I use eric6 for python and the latest version requires sip >= 5.0. I understand that sip after 4.x no longer has standalone releases. Would it be possible to upgrade it to the latest version, either at about this time or after release? I think source can be grabbed from this link (see the last file, a tarball) https://pypi.org/project/sip/#files Thanks! |
|
p/mariadb-10.6.0-x86_64-1.txz WTF?
"The MariaDB Foundation is pleased to announce the availability of MariaDB 10.6.0, the first alpha release in the new MariaDB 10.6 development series." |
Quote:
Code:
Mon Apr 12 20:07:12 UTC 2021Code:
This update fixes the illegal instruction regession on 32-bit with processors |
Could we perhaps have gcc built with support for JIT*, in anticipation of Emacs 28? :)
*There's some packaging notes on that page about having to build twice with and without '--enable-host-shared' (basically once for JIT with --enable-host-shared and --enable-languages=jit, and once with just --enable-languages=all). |
Quote:
Perhaps, although 10.5.10 is imminent in the stable tree, I have a copy, so I expect it publicly released very, very soon ;) and well I guess -current still technically is a pre production with no guarantees, we all know the risks. |
I was waiting for it too but unfortunately it hasn't even been tagged yet in their git
https://github.com/MariaDB/server/releases |
adwaita-icon-theme-40.1
https://download.gnome.org/sources/a...me-40.1.tar.xz |
Is the final conformation of the "Basic Network Setup" in the installer really needed?
Just seems like a extra step for nothing really. |
Wireguard, openresolv
Hi,
I tested wg-quick with and without openresolv. It will work properly in both cases, provided that the client configuration does not request setting up DNS in /etc/resolv.conf, with the DNS = X.X.X.X directive. That is, wg-quick uses openresolv to add entry to /etc/resolv.conf if the client configuration file has DNS directive. I have nothing against adding openresolv to Slackware. That being said, if it's added, then IMO NetworkManager should be configured to use it by default. This should avoid issues should the users use openresolv and NM at the same time. I did a bit of testing and indeed when bringing up WG interface, the DNS from WG config file replaces the ones from /etc/resolv.conf. When the WG interface is torn down, the previous entry in /etc/resolv.conf is restored. Whether this is exactly the expected behaviour remains to be discussed. I use NM to configure the main interface. On a side note, NM has native support for Wireguard. -- Best regards, Andrzej Telszewski |
Quote:
Code:
testing/packages/mariadb-10.6.0-x86_64-1.txz: Upgraded. |
As noted in the dedicated thread [ https://www.linuxquestions.org/quest...cs-4175694634/ ] future Emacs will require libgccjit. It would be nice to have it now [ https://gcc.gnu.org/onlinedocs/jit/i...he-jit-library ]
|
Please undo distribution of httpd 2.4.47. It contains a regression bug and was pulled from the release after the vote.
https://lists.apache.org/x/thread.ht....apache.org%3E Quote:
|
Quote:
-current runs fine with this version, and it has not been pulled, its still available on mirrors, just not being announced. 2.4.48 will be released (if passes) on May 23 or 24 |
I like SCHED_AUTOGROUP
Quote:
find arch/*/configs/ -type f -exec grep -F CONFIG_SCHED_AUTOGROUP {} + I have a FX6300/990FX 16GB PC, my root and home are on a SSD but I also have a WD80EZAZ where I store backups and big disk images for my VM's. It's still a decent system and I use it daily. I hope ASMedia now sells better SATA silicon to AMD, but that's what I have to deal with now. During heavy disk I/O on the rotating disk (e.g. backing up big files) Mate DE becomes irresponsive, even if the CPU's are mostly idle. Qemu doesn't care too much, but VirtualBox guests that do a lot of I/O are always unusable (I can't imagine running them during a backup). A few days ago I decided to give autogroup a try. VirtualBox is still definitely slower than Qemu, but it went from unusable to slow, and this is a colossal improvement for me. Heavy rotating disk I/O also no longer hangs my DE. I got used pretty quickly to this new smoothness. And I like it to the point that I can now easily notice when autogroup is disabled or not. Here is a silly benchmark I just made: cd /dev/shm/linux-5.10.34 && make mrproper defconfig && time make -j6 all autogroup on: real 6m56,369s user 36m46,422s sys 3m19,560s autogroup off: real 6m54,400s user 36m43,876s sys 3m13,808s It is a half percent penalty ("real" is 6m 48s slower every 24 hours of kernel compilation). Unless you mine crypto currencies with your CPU, I think autogroup is desirable. GazL is right: being able to disable it is much better than not having it at all. |
Quote:
Code:
scsi_mod.use_blk_mq=1Code:
ACTION=="add|change", KERNEL=="sd[a-z]|mmcblk[0-9]*|nvme[0-9]*", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="mq-deadline" |
Quote:
P.S. I always run intensive background workloads under: chrt -i 0 ionice -c3 Keeps everything nice and responsive. |
adwaita-icon-theme-40.1.1
https://download.gnome.org/sources/a...-40.1.1.tar.xz |
1 Attachment(s)
Quote:
I tried all the schedulers with the crude script I attached. You were right, "bfq" and "none" (who knew!) on my PC are the fastest by a good margin: Code:
find cp umount totalThank you for investing your time on my ~10 years old mobo :) |
Quote:
(BTW, when I do benchmarks, I use 'echo 3 > /proc/sys/vm/drop_caches' just before timing.) |
Quote:
The original patch came with benchmark results. There were eleven CPU bound processes running on ONE core: a kernel was built with -j10 and simultaneously another process in a loop only got one eleventh (9%) of cpu time. With autogroup the other process gets 50%. Why would anyone 'make -j10' on one core? With 'make -j1' the other process would have got 50%. And 'nice' would give even more. But with autogroup nice does not help. What if you are building a kernel (-j1) in one xterm and simultanously you are building qt (-j1) in another xterm (using one-core cpu)? They are in different groups because they were launched in different terminals. And all the processes you start from your WM belong to one single group. So, if you have three interactive processes: firefox looping in javascript, chromium looping in javascript, and you are watching a video, autogroup gives them all only 33% cpu to be divided between them, 33 % to the kernel building process and 33% to the qt building process. Nice does not help. Without autogroup, the three interactive processes firefox+chromium+video would get 60%, the kernel build 20%, and the qt build 20%. And nice would help. Same thing with the Phoronix video/benchmark: they showed autogroup using 'make -j64', with a 6-core cpu... |
|
Mozilla Firefox 88.0.1
https://ftp.mozilla.org/pub/firefox/....source.tar.xz Mozilla Thunderbird 78.10.1 https://ftp.mozilla.org/pub/thunderb....source.tar.xz Mutt 2.0.7 ftp://ftp.mutt.org/pub/mutt/mutt-2.0.7.tar.gz |
Mesa 21.1.0 is available (feature release).
|
Is it possible to add libopenmpt (https://lib.openmpt.org/libopenmpt/) and recompile audacious-plugins to use it. As it sometimes useful to have module files support out of the box for playing module files like .xm, .it and etc. Audacious works quite well with libopenmpt.
|
If it's possible to add ifuse to mount storage and access Apple iDevice documents ?
https://github.com/libimobiledevice/ifuse |
|
|
|
Quote:
Until now, only the systemd had support for CGROUP v2 and everyone else was stuck on CGROUP v1. But, now there's hope that it will be used also by elogind, LXC and any other software with interests on CGROUP's noble art. |
mariadb 10.5.10 is now released
|
and since I dont see it mentioned, back on Thursday PHP 7.4.19 was released
I dont use pgsql but it apparently reverts a bug related to PDO_pgsql that was introduced in PHP 7.4.18. |
Quote:
Code:
Wed May 5 19:56:53 UTC 2021 |
| All times are GMT -5. The time now is 03:23 PM. |