[arch-releng] why this was removed from archiso hook?
Hello Pierre, https://projects.archlinux.org/archiso.git/commit/?id=76f275e993eec849bd7c03... - echo "/dev/mapper/${dm_snap_name} ${mnt} auto defaults 0 0" >> ${newroot}/etc/fstab looks like an unrelated change. (anyway looks like this does not make sense with systemd, and system boots fine without errors) And this other https://projects.archlinux.org/archiso.git/commit/?id=d355c2b83527d3f0c08c6e... * the multi-user.target is the default is not the default! graphical.target is the default, this causes an error on startup because is trying to start display-manager.service, so fallback to multi-user.target Thanks. -- Gerardo Exequiel Pozzi \cos^2\alpha + \sin^2\alpha = 1
Am 07.10.2012 00:33, schrieb Gerardo Exequiel Pozzi:
Hello Pierre,
https://projects.archlinux.org/archiso.git/commit/?id=76f275e993eec849bd7c03...
- echo "/dev/mapper/${dm_snap_name} ${mnt} auto defaults 0 0" >> ${newroot}/etc/fstab
looks like an unrelated change. (anyway looks like this does not make sense with systemd, and system boots fine without errors)
The .mount units created by these fstab entries are noops. Furthermore, the fstab entries cause systemd-remount-fs to be called and potentially do something (but we definitely don't want that). Adding anything to fstab has no benefit, but the potential to cause problems.
And this other
https://projects.archlinux.org/archiso.git/commit/?id=d355c2b83527d3f0c08c6e...
OMG, this is shit. This needs to be (mostly) reverted. This replaces the normal getty@.service by our autologin@.service, which causes autologin on all terminals. This is definitely not intended. We don't want to replace getty@.service with autologin, we only want to start autologin on tty1.
* the multi-user.target is the default
is not the default! graphical.target is the default, this causes an error on startup because is trying to start display-manager.service, so fallback to multi-user.target
Correct. Pierre, why are you doing last minute changes to things you clearly do not understand properly?
The only valid part of this patch was the removal of haveged.service, the rest was incorrect. We should NOT override getty@.service with autologin@.service, instead the autologin@.service is meant as a separate unit. In contrast to the commit message, multi-user.target is not the default default target and still has to be enabled. --- configs/releng/build.sh | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/configs/releng/build.sh b/configs/releng/build.sh index 2699468..b78d2b0 100755 --- a/configs/releng/build.sh +++ b/configs/releng/build.sh @@ -183,7 +183,7 @@ make_customize_root_image() { patch ${work_dir}/root-image/usr/bin/pacman-key < ${script_path}/pacman-key-4.0.3_unattended-keyring-init.patch sed -i 's/#\(en_US\.UTF-8\)/\1/' ${work_dir}/root-image/etc/locale.gen sed 's#\(^ExecStart=-/sbin/agetty\)#\1 --autologin root#' \ - ${work_dir}/root-image/usr/lib/systemd/system/getty@.service > ${work_dir}/root-image/etc/systemd/system/getty@.service + ${work_dir}/root-image/usr/lib/systemd/system/getty@.service > ${work_dir}/root-image/etc/systemd/system/autologin@.service mkarchiso ${verbose} -w "${work_dir}" -C "${pacman_conf}" -D "${install_dir}" \ -r 'locale-gen' \ run @@ -194,7 +194,7 @@ make_customize_root_image() { -r 'useradd -m -p "" -g users -G "adm,audio,floppy,log,network,rfkill,scanner,storage,optical,power,wheel" -s /bin/zsh arch' \ run mkarchiso ${verbose} -w "${work_dir}" -C "${pacman_conf}" -D "${install_dir}" \ - -r 'systemctl -f enable pacman-init.service getty@.service dhcpcd.service || true' \ + -r 'systemctl -f enable multi-user.target pacman-init.service autologin@.service dhcpcd.service || true' \ run : > ${work_dir}/build.${FUNCNAME} fi -- 1.7.12.2
Am 07.10.2012 00:52, schrieb Thomas Bächler:
https://projects.archlinux.org/archiso.git/commit/?id=d355c2b83527d3f0c08c6e...
OMG, this is shit. This needs to be (mostly) reverted.
This replaces the normal getty@.service by our autologin@.service, which causes autologin on all terminals. This is definitely not intended.
We don't want to replace getty@.service with autologin, we only want to start autologin on tty1.
This is exactly what it is doing here. How did you manage to have it logged in on all terminals? I have tested the iso several times on vm and different real hardware and never had this issue.
* the multi-user.target is the default
is not the default! graphical.target is the default, this causes an error on startup because is trying to start display-manager.service, so fallback to multi-user.target
Correct.
Sorry, I forgot about the display-manager issue (afaik this was introduced in later versions of systemd). But you guys are a little overreacting here. I don't see how this would justify a new iso release. All that happens is that systemd prints a warning in its journal. There is no error on startup; it even prints a green OK here. It is correct to enable the multi-user.target again. But to be clear here, there is no visible effect as this target was activated by the graphical target anyway.
Pierre, why are you doing last minute changes to things you clearly do not understand properly?
How about you watch your tone? Nobody is going to work on this project if we insult each other when minor errors are made. Greetings, Pierre -- Pierre Schmitz, https://pierre-schmitz.com
Am 07.10.2012 10:30, schrieb Pierre Schmitz:
We don't want to replace getty@.service with autologin, we only want to start autologin on tty1.
This is exactly what it is doing here. How did you manage to have it logged in on all terminals? I have tested the iso several times on vm and different real hardware and never had this issue.
I just read the code, I didn't make an ISO. And the code said that you replaced getty@.service with what used to be autologin@.service. This would still work as expected, but for a very weird reason: While you override getty@.service, autovt@.service is still the same, because it is symlinked to the /usr/lib version of getty@.service. This is extremely unintuitive in my opinion. However, you are right that it wouldn't break behaviour, it's just ugly.
Pierre, why are you doing last minute changes to things you clearly do not understand properly?
How about you watch your tone? Nobody is going to work on this project if we insult each other when minor errors are made.
This is about last-minute changes. You commited changes to the repository without sending them through this mailing list first. This is not how we should work.
Am 07.10.2012 10:59, schrieb Thomas Bächler:
This is about last-minute changes. You commited changes to the repository without sending them through this mailing list first. This is not how we should work.
I regret doing that. Our current work flow is not ideal. We did some major changes right before a new iso should have been built. I'll send a new mail with a proposal how we could learn from this. Greetings, Pierre -- Pierre Schmitz, https://pierre-schmitz.com
Am 07.10.2012 12:09, schrieb Pierre Schmitz:
Am 07.10.2012 10:59, schrieb Thomas Bächler:
This is about last-minute changes. You commited changes to the repository without sending them through this mailing list first. This is not how we should work.
I regret doing that. Our current work flow is not ideal. We did some major changes right before a new iso should have been built. I'll send a new mail with a proposal how we could learn from this.
I should also fix the 'git push' emails.
On 10/07/2012 05:30 AM, Pierre Schmitz wrote:
Am 07.10.2012 00:52, schrieb Thomas Bächler:
https://projects.archlinux.org/archiso.git/commit/?id=d355c2b83527d3f0c08c6e... OMG, this is shit. This needs to be (mostly) reverted.
This replaces the normal getty@.service by our autologin@.service, which causes autologin on all terminals. This is definitely not intended.
We don't want to replace getty@.service with autologin, we only want to start autologin on tty1. This is exactly what it is doing here. How did you manage to have it logged in on all terminals? I have tested the iso several times on vm and different real hardware and never had this issue.
* the multi-user.target is the default
is not the default! graphical.target is the default, this causes an error on startup because is trying to start display-manager.service, so fallback to multi-user.target Correct. Sorry, I forgot about the display-manager issue (afaik this was introduced in later versions of systemd). But you guys are a little overreacting here. I don't see how this would justify a new iso release. All that happens is that systemd prints a warning in its journal. There is no error on startup; it even prints a green OK here.
It is correct to enable the multi-user.target again.
But to be clear here, there is no visible effect as this target was activated by the graphical target anyway.
Pierre, why are you doing last minute changes to things you clearly do not understand properly? How about you watch your tone? Nobody is going to work on this project if we insult each other when minor errors are made.
Greetings,
Pierre
sorry if I sound in a bad tone, this was not my intention. so, what to do revert or not revert? -- Gerardo Exequiel Pozzi \cos^2\alpha + \sin^2\alpha = 1
On 10/08/2012 05:08 AM, Gerardo Exequiel Pozzi wrote:
On 10/07/2012 05:30 AM, Pierre Schmitz wrote:
Am 07.10.2012 00:52, schrieb Thomas Bächler:
https://projects.archlinux.org/archiso.git/commit/?id=d355c2b83527d3f0c08c6e...
OMG, this is shit. This needs to be (mostly) reverted.
This replaces the normal getty@.service by our autologin@.service, which causes autologin on all terminals. This is definitely not intended.
We don't want to replace getty@.service with autologin, we only want to start autologin on tty1. This is exactly what it is doing here. How did you manage to have it logged in on all terminals? I have tested the iso several times on vm and different real hardware and never had this issue.
* the multi-user.target is the default
is not the default! graphical.target is the default, this causes an error on startup because is trying to start display-manager.service, so fallback to multi-user.target Correct. Sorry, I forgot about the display-manager issue (afaik this was introduced in later versions of systemd). But you guys are a little overreacting here. I don't see how this would justify a new iso release. All that happens is that systemd prints a warning in its journal. There is no error on startup; it even prints a green OK here.
It is correct to enable the multi-user.target again.
But to be clear here, there is no visible effect as this target was activated by the graphical target anyway.
Pierre, why are you doing last minute changes to things you clearly do not understand properly? How about you watch your tone? Nobody is going to work on this project if we insult each other when minor errors are made.
Greetings,
Pierre
sorry if I sound in a bad tone, this was not my intention.
so, what to do revert or not revert?
anybody? getty.service name should be reverted to autologin? -- Gerardo Exequiel Pozzi \cos^2\alpha + \sin^2\alpha = 1
Am 19.10.2012 03:05, schrieb Gerardo Exequiel Pozzi:
sorry if I sound in a bad tone, this was not my intention.
so, what to do revert or not revert?
anybody? getty.service name should be reverted to autologin?
IMO, yes. To make this entirely clean, we should also explicitly disable getty@tty1.service and change the [Install] section of autologin@.service to name the symlink properly (as in, autologin@tty1.service).
-f is not needed anymore, disable explicitly default getty service Signed-off-by: Gerardo Exequiel Pozzi <vmlinuz386@yahoo.com.ar> --- configs/releng/build.sh | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/configs/releng/build.sh b/configs/releng/build.sh index e758ff3..006f3fb 100755 --- a/configs/releng/build.sh +++ b/configs/releng/build.sh @@ -174,8 +174,9 @@ make_customize_root_image() { sed -i "s/#Server/Server/g" ${work_dir}/root-image/etc/pacman.d/mirrorlist patch ${work_dir}/root-image/usr/bin/pacman-key < ${script_path}/pacman-key-4.0.3_unattended-keyring-init.patch sed -i 's/#\(en_US\.UTF-8\)/\1/' ${work_dir}/root-image/etc/locale.gen - sed 's#\(^ExecStart=-/sbin/agetty\)#\1 --autologin root#' \ - ${work_dir}/root-image/usr/lib/systemd/system/getty@.service > ${work_dir}/root-image/etc/systemd/system/getty@.service + sed 's#\(^ExecStart=-/sbin/agetty\)#\1 --autologin root#; + s#\(^Alias=getty.target.wants/\).\+#\1autologin@tty1.service#' \ + ${work_dir}/root-image/usr/lib/systemd/system/getty@.service > ${work_dir}/root-image/etc/systemd/system/autologin@.service mkarchiso ${verbose} -w "${work_dir}" -C "${pacman_conf}" -D "${install_dir}" \ -r 'locale-gen' \ run @@ -186,7 +187,10 @@ make_customize_root_image() { -r 'useradd -m -p "" -g users -G "adm,audio,floppy,log,network,rfkill,scanner,storage,optical,power,wheel" -s /bin/zsh arch' \ run mkarchiso ${verbose} -w "${work_dir}" -C "${pacman_conf}" -D "${install_dir}" \ - -r 'systemctl -f enable multi-user.target pacman-init.service getty@.service dhcpcd.service || true' \ + -r 'systemctl disable getty@tty1.service || true' \ + run + mkarchiso ${verbose} -w "${work_dir}" -C "${pacman_conf}" -D "${install_dir}" \ + -r 'systemctl enable multi-user.target pacman-init.service autologin@.service dhcpcd.service || true' \ run : > ${work_dir}/build.${FUNCNAME} fi -- 1.7.12.3
Am 20.10.2012 18:47, schrieb Gerardo Exequiel Pozzi:
-f is not needed anymore, disable explicitly default getty service
Signed-off-by: Gerardo Exequiel Pozzi <vmlinuz386@yahoo.com.ar>
Didn't test this, but from the code it seems exactly right. +1.
participants (3)
-
Gerardo Exequiel Pozzi
-
Pierre Schmitz
-
Thomas Bächler