Requests for -current (14.2-->15.0)

zhl · 03-22-2021, 06:50 PM

Quote:

Originally Posted by jostber

Add an option to pkgtool and setup that is called Repair which runs these commands in one - Find bootloader installed, Run geninitrd, Run bootloader command, Output kernel version and bootload conf file at the end.

Another request to pkgtools for repairing slackpkg if database corruptions, my last reinstallation caused by a bug of removepkg dir endwith wildcase(*).

shastah · 03-22-2021, 09:29 PM

Quote:

Originally Posted by dgusev

Suggest to switch installation package default compression algorithm to zstandard.

I concur.

Pkgtools advertises "--threads <number>" for decompression, but in real life it's not very useful, as all vanilla Slackware packages are compressed with XZ, and XZ does not support parallel decompression (only compression).

Using t/texlive-2020.200608-x86_64-4.txz as an example, decompressing it on some i5 Intel NUC I have lying around takes 6.8s. After repackaging it with zstd (xz -dc | zstd -19 -T0 -c), decompressing with zstd then takes 0.6s.
That's more than ten times speedup, for a price of a ~9% of space (76MB xz, 83M zstd).
And remember that installpkg goes through the archive at least two times (sometimes even three).

Lockywolf · 03-22-2021, 11:11 PM

Following the discussion here: https://gitlab.xfce.org/apps/xfce4-s...er/-/issues/38

I would like to ask, if possible, to add a run of the script (https://gist.githubusercontent.com/b...con_patcher.py) to each theme in gnome-themes-extra.SlackBuild

Otherwise, most themes from there are instantly becoming inapplicable to xfce. Since it is mostly creating symlinks, maybe it should be added to the doinst.sh?

y0g1 · 03-23-2021, 02:48 AM

My request for -current:
I would like to change /etc/rc.d/rc.lxc to start_lxc and stop_lxc in background, then wait for them to finish actions.
something like:
for start_lxc:
- start all needed containers without waiting for result
- wait for results all started containers
for stop_lxc:
- stop all running containers without waiting for result (-W option)
- wait for results all stopped containers

Code:

start_lxc() {
  LXC_TO_PROCESS=$(/usr/bin/lxc-ls)
  LXC_TO_WAIT=""
  for CONTAIN in ${LXC_TO_PROCESS}; do
    if [ "$(lxc-info -n ${CONTAIN} -c lxc.start.auto)" = "lxc.start.auto = 1"  ]; then
      if [ "$(/usr/bin/lxc-info -s -n ${CONTAIN} | grep STOPPED$)" ]; then
        echo "Starting LXC container ${CONTAIN}."
        LXC_TO_WAIT="${LXC_TO_WAIT} ${CONTAIN}"
        /usr/bin/lxc-start -n ${CONTAIN}
      fi
    fi
  done
  for CONTAIN in ${LXC_TO_WAIT}; do
    /usr/bin/lxc-wait -n ${CONTAIN} -s RUNNING
    if [ $? -gt 0 ]; then
      return 2
    fi
  done
}

stop_lxc() { 
  LXC_TO_PROCESS=$(/usr/bin/lxc-ls --active) 
  for CONTAIN in ${LXC_TO_PROCESS}; do 
    echo "Stopping LXC container ${CONTAIN}." 
    /usr/bin/lxc-stop -W -n ${CONTAIN} 
  done 
  for CONTAIN in ${LXC_TO_PROCESS}; do 
    echo "Waiting for LXC container ${CONTAIN}." 
    /usr/bin/lxc-wait -n ${CONTAIN} -s STOPPED 
    if [ $? -gt 0 ]; then 
      return 2 
    fi 
  done 
}

What do you think about it?

walecha · 03-23-2021, 03:07 AM

Quote:

Originally Posted by dgusev

Suggest to switch installation package default compression algorithm to zstandard.
initramfs image also could be compressed with zstandard by default, Slackware generic kernel already has zstandard for initramfs support enabled (CONFIG_RD_ZSTD=y)

I'm 100% with this since I'm already using zstd to zip my initrd since last year by editing /sbin/mkinitrd to use zstdmt instead of gzip.

avian · 03-23-2021, 04:55 AM

Quote:

Originally Posted by walecha

I'm 100% with this since I'm already using zstd to zip my initrd since last year by editing /sbin/mkinitrd to use zstdmt instead of gzip.

Same. Just this week I moved to zstd compression for the kernel and zstd backups through clonezilla. So many positives, so few negatives. Can only imagine how quick a slackware install would be if packages moved to zstd.

Tonus · 03-23-2021, 05:10 PM

Please let's deal with dialogrc as a .new config file. I now have a nice theme and get it overridden when dialog gets updates (usually forget to fix it before using some curse interface).

01micko · 03-24-2021, 07:02 AM

Please include the latest epson-escpr driver from seiko-epson (it is GPL licensed). The one included in CUPS is buggy because all it does is feed paper though after trying to print the test page; a hard reset of the printer is needed to stop it! I'm on a fresh download and update of -current 64 as of half an hour ago.

My printer is an Epson Expression XP-245.

If I install the package made from the Slackbuild from https://github.com/Ponce/slackbuilds...-printer-escpr (after deleting the printer) all is well

Thank you.

opty · 03-24-2021, 09:16 AM

Please remove postdrop user from CHANGES_AND_HINTS.TXT and ChangeLog.txt: Only postdrop group exists.

Andypoo · 03-24-2021, 07:29 PM

Just reposting this at it hasn't been picked up yet. I think this would be a great inclusion as it fixes a long-standing issue with making the initrd.gz for MMC systems, which causes a fail-to-find-rootfs kernel panic on next boot otherwise.

Quote:

Originally Posted by Andypoo

mmc_block module is still getting missed from the initial initrd.img that gets created when doing installs on mmcblk0 devices.

I dug into this further. The issue stems from the mmc_block module having an internal name of mmcblk. This means that the mkinitrd_command_generator.sh script in mkinitrd to generate the module list misses it.

I put in a specific workaround for this in the patch below. A little ugly, but gets the job done. And can't think of a much better way given the mismatch in the module's module and internal name.

Andrew.

Code:

--- mkinitrd_command_generator.sh	2021-03-15 18:15:39.000000000 +0200
+++ mkinitrd_command_generator_new.sh	2021-03-15 18:15:29.000000000 +0200
@@ -212,6 +212,7 @@
   MLIST=$(for i in $(find /sys/block/*/ -name "device" -print0 | xargs -0 -i'{}' readlink -f '{}' | sort -u); do
     /sbin/udevadm info --query=all --path=$i --attribute-walk | \
       sed -ne 's/^[[:blank:]]\+DRIVER[S]*=="\([^"]\+\)"$/\1/p' | \
+      sed -e 's/^mmcblk$/mmc_block/' | \
       xargs -I@ /sbin/modprobe --set-version $KVER --show-depends @ \
       2>/dev/null | grep -v "builtin " | \
       while read LINE ; do

biker_rat · 03-24-2021, 07:54 PM

Mesa 21.0.1

Ser Olmy · 03-24-2021, 09:28 PM

I just ran slackpkg upgrade-all on two of my 64-current systems, and rc.inet1 appears to be extremely broken.

I have a single interface configured with USE_DHCP[n]="yes" (eth1 on one box and eth4 on the other), and after waiting less than 5 seconds it fails to obtain an address and configures APIPA incorrectly instead. Adding a DHCP_TIMEOUT[n] setting does not help, as the setting is just ignored.

There's no gateway setting in rc.inet1.conf, but a weird, invalid default interface route is still added to the routing table. As a result, absolutely no IP connectivity is available, including APIPA. I have to kill dhcpcd and manually delete the invalid route/IP address, and only then will dhcpcd or ip addr add work.

I did originally use an old config file, but regenerating rc.inet1.conf with netconfig did not fix the issue.

saxa · 03-25-2021, 09:02 AM

dconf-editor-3.38.3
https://download.gnome.org/sources/d...-3.38.3.tar.xz

cwizardone · 03-25-2021, 11:28 AM

"OpenSSL 1.1.1k is now available, including bug and security fixes."
The tarball, ftp://ftp.openssl.org/source/openssl-1.1.1k.tar.gz

The OpenSSL web site,https://www.openssl.org/

tadgy · 03-26-2021, 12:05 AM

Quote:

Originally Posted by Ser Olmy

I just ran slackpkg upgrade-all on two of my 64-current systems, and rc.inet1 appears to be extremely broken.

I have a single interface configured with USE_DHCP[n]="yes" (eth1 on one box and eth4 on the other), and after waiting less than 5 seconds it fails to obtain an address and configures APIPA incorrectly instead. Adding a DHCP_TIMEOUT[n] setting does not help, as the setting is just ignored.

There's no gateway setting in rc.inet1.conf, but a weird, invalid default interface route is still added to the routing table. As a result, absolutely no IP connectivity is available, including APIPA. I have to kill dhcpcd and manually delete the invalid route/IP address, and only then will dhcpcd or ip addr add work.

I did originally use an old config file, but regenerating rc.inet1.conf with netconfig did not fix the issue.

Thanks for the feedback; hopefully we can get to the root of the trouble.

Firstly, can I ask you to pop over to the thread titled "Request for testing: new rc.inet1" as that's a more appropriate place to get into debugging.

Additionally, can you please set

Code:

DEBUG_ETH_UP="yes"

and do an

Code:

/etc/rc.d/rc.inet1 eth1_restart

This should produce a full log of what commands were executed to bring the interface up in syslog - please can you post that to the other thread so I can take a look.

Please also post your rc.inet1.conf section for your eth1 (or eth4) interface - are you seeing anything other than USE_DHCP[x] ?

My suspicion is that this is a DHCP client->server issue and not related to rc.inet1 at all - once it calls dhcpcd with the correct arguments, all bets are off.

See you over at the other thread