Ascii uninstallable, multiple other issues encountered during install
-
Ascii is currently not installable. initramfs-tools breaks the current initscripts version, installer fails installing the base system when it tries installing initramfs-tools at the kernel step.
-
This bug tracker required me to sign up for an account, with password security requirements, and email verification shit, just to report an issue. Do you want bug reports or not? It also is ugly, uses little icons for navigation, and has the Submit button on the far left. It also uses a static bar over the top, that breaks scrolling a page at a time, either by the keyboard or scroll bar.
-
apt-get reports an utterly incorrect error about the issue. apt-get says initramfs-tools depends on initramfs-tools-core but it's not being installed, for absolutely no reason, and even as it happily installs initramfs-tools-core. The actual error has to do with initscripts breaks. This appears to be an apt-get bug. Aptitude correctly reports the issue. Here's an example: http://imgur.com/gLJ7TaR
-
One of the files failed to download due to network issues. The error says to check the fourth console and /var/log/syslog. However, absolutely nothing about the failed download is reported to either of these. The downloader needs to output information about what failed and why, not just silently return an error.
-
Failed downloaded files should be retried without having to restart the entire base system install process.
-
If you do something that annoys the installer, reopen-console and steal-ctty are used to restart it. But they don't work. The installer starts, but is non-interactive, in that you can't interact with it normally. It seems that things you input are either acted on later when you're not looking, or acted on without the console being updated. This could result in you selecting dangerous options by accident, and at a minimum prevents the install from completing. Not much point in restarting the installer if you can't actually use it.
-
reopen-console segvs when called with no arguments
-
Attempting to restart the installer needs a rate limit or a retry limit. If it fails to restart, it results in a rapidly flashing white screen, that you have to ignore while fixing it from the console.
-
If it's bad to use security.debian.org currently, don't have it selected by default. Best option may be to create an empty security.devuan.org repo and select that by default, so users end up with updates automagically once devuan security is up and running.
-
Packages, release, or other files are re-downloaded every time the install the base system step is started. This wastes lots of bandwidth and time, if you're not on a fast connection. Should check if there's actually a newer version before redownloading them.
-
The download status when installing the kernel mis-calculates the number of files to download. It goes 1 of 2, 2 of 2, 3 of 3, 4 of 4. Very minor issue, of course...
-
Network settings selected during install should be copied over to the install, especially wifi. Setting up wpa_supplicant manually is a bit annoying, and will be very annoying for a new user with no idea what's wrong and no internet access to read up on why.
-
This is a long-term issue with apt, but it sure would be nice if apt-get didn't entirely fail any time there's any issue at all with any installed package dependencies or conflicts, and refuse to do anything with any package even if entirely unrelated. You should be able to --force something without the entire package management system becoming useless.
-
The message about where to install EFI stuff isn't clear enough. It's a try, but many people won't even know what EFI is. Needs more Just Works, or clearer help.
There were more, but I'm writing this after, not during, and they've already fallen out. Oh well. Make the issue system require less hoop-jumping and the next user installing can report them. Many of these probably just need to be passed on to upstream maintainers anyway.
Keep up the good work towards software freedom!
-
Hi @Bushytails
Thanks for the bug report. Addressing your points:
- Ascii is currently not installable. initramfs-tools breaks the current initscripts version, installer fails installing the base system when it tries installing initramfs-tools at the kernel step.
Ok, this looks like an upstream issue as we don't build our own package for initramfs-tools (yet).
- This bug tracker required me to sign up for an account, with password security requirements, and email verification shit, just to report an issue. Do you want bug reports or not? It also is ugly, uses little icons for navigation, and has the Submit button on the far left. It also uses a static bar over the top, that breaks scrolling a page at a time, either by the keyboard or scroll bar.
Unfortunately that's the nature of things these days. Either make properly authenticated signups a requirement or have to cope with huge amounts of spam. We plan to have our own version of reportbug setup soon that will allow you to file reports from the running system, and I would dearly love to be able to build into the installer an installation report feature that can be used to collect install details and make reporting installation failures and successes and suggestions much easier to do.
- apt-get reports an utterly incorrect error about the issue. apt-get says initramfs-tools depends on initramfs-tools-core but it's not being installed, for absolutely no reason, and even as it happily installs initramfs-tools-core. The actual error has to do with initscripts breaks. This appears to be an apt-get bug. Aptitude correctly reports the issue. Here's an example: http://imgur.com/gLJ7TaR
Wow, that's good to know. How did you come to that conclusion. Were you running apt-get from within the installed system? I'll dig into this later and see if we need to file a bug upstream.
- One of the files failed to download due to network issues. The error says to check the fourth console and /var/log/syslog. However, absolutely nothing about the failed download is reported to either of these. The downloader needs to output information about what failed and why, not just silently return an error.
It's probable that the failed download was early in the list and you didn't scroll back far enough.
- Failed downloaded files should be retried without having to restart the entire base system install process.
As far as I'm aware failed downloads are retried at least once, but I'll have to double check this. Of course if this is a mirror such as missing package then re-trying the download is likely to fail anyway.
- If you do something that annoys the installer, reopen-console and steal-ctty are used to restart it. But they don't work. The installer starts, but is non-interactive, in that you can't interact with it normally. It seems that things you input are either acted on later when you're not looking, or acted on without the console being updated. This could result in you selecting dangerous options by accident, and at a minimum prevents the install from completing. Not much point in restarting the installer if you can't actually use it.
Hmmm... That to me indicates that the previous installer ended in an undefined state, or it didn't die.
- reopen-console segvs when called with no arguments
- Attempting to restart the installer needs a rate limit or a retry limit. If it fails to restart, it results in a rapidly flashing white screen, that you have to ignore while fixing it from the console.
- If it's bad to use security.debian.org currently, don't have it selected by default. Best option may be to create an empty security.devuan.org repo and select that by default, so users end up with updates automagically once devuan security is up and running.
Good point. We will be setting up the devuan security repo soon. Probably around the time we push the beta installer out. I have been looking into this.
- Packages, release, or other files are re-downloaded every time the install the base system step is started. This wastes lots of bandwidth and time, if you're not on a fast connection. Should check if there's actually a newer version before re-downloading them.
Good point. Ideally we don't want the the base system-install to be restarted, but re-downloading already downloaded packages isn't good either.
- The download status when installing the kernel mis-calculates the number of files to download. It goes 1 of 2, 2 of 2, 3 of 3, 4 of 4. Very minor issue, of course...
Good spotting. I'll have to dig around and find out what package is responsible for this issue.
- Network settings selected during install should be copied over to the install, especially wifi. Setting up wpa_supplicant manually is a bit annoying, and will be very annoying for a new user with no idea what's wrong and no internet access to read up on why.
Agree'd, but that is hard to do without knowing in advance which tool will be used for managing the wireless interface in the installed system.
- This is a long-term issue with apt, but it sure would be nice if apt-get didn't entirely fail any time there's any issue at all with any installed package dependencies or conflicts, and refuse to do anything with any package even if entirely unrelated. You should be able to --force something without the entire package management system becoming useless.
Agree'd, but I think that is more a structural issue within either apt or dpkg...
- The message about where to install EFI stuff isn't clear enough. It's a try, but many people won't even know what EFI is. Needs more Just Works, or clearer help.
If you can suggest wording. I'll look at what I can do to improve that.
There were more, but I'm writing this after, not during, and they've already fallen out. Oh well. Make the issue system require less hoop-jumping and the next user installing can report them. Many of these probably just need to be passed on to upstream maintainers anyway.
Keep filing bugs and if you're inclined try to propose patches or at least suggestions about causes and resolutions
Keep up the good work towards software freedom!
And thank you again for a thorough bug report and your support of our efforts.
Edited by Daniel Reurich -
looks like this may be the source of the issue:
-
... OK, is there some bug with the bug tracker? I'm trying to reply, I click the green button, and... nothing happens. Is this being posted ten times? I have no fucking idea. The button changes to light green and shows a no symbol, then a couple seconds later changes back to dark green, and nothing's changed. I guess it's not being posted ten times, as loading the issue in another tab doesn't show it at all. I'll try again later. Ah, now it's showing that I'm logged out. I'll try logging in from another tab, as I sure don't want to re-type all this. Nope, that wasn't it either. I'll try copy-pasting it to the reply box on another tab.
No, that bug has already been addressed by the installer directly installing busybox prior to installing initramfs-tools. The issue installing initramfs-tools is due to it having a breaks: for initscripts of the only available version.
-
just needs initscripts' version number bumped most likely, with whatever supposedly required change made. Devuan seems to have a slightly old version of initscripts packaged, which the current initramfs-tools is marked as breaking.
-
I was running apt-get and aptitude from within the mostly-installed base system. The error on console 4 was about initramfs-tools-core not going to be installed, so I used one of the other consoles, chrooted to the target system, and was playing around with it to see why. apt-get would always complain it was due to initramfs-tools-core not being installable, even as it would happily install it at the same time, or it was already installed. Check out the old-fashioned screenshot I got, where apt-get is complaining about it even when it's already installed. I installed aptitude and it identified the correct issue.
-
there was nothing to scroll. Nothing about downloads was outputted to console 4 at all. And if you can scroll it, I don't know how, as the usual shift-pgup does nothing...
-
I was assuming it was my internet connection, as it's unreliable, and it succeeded when I tried again. I guess it could have been missing from the repo temporarily.
-
yeah, I pissed the installer off and it crashed, in a way that's not due to an installer bug... The bug is it being restarted but with some console input/output issue. Hopefully most users don't encounter installer crashes and will never see it, but if it's going to have an automatic restart, it should work. :)
-
I'm on a 3g hotspot with a weak signal and $11/gb bandwidth fees... downloading those files repeatedly was a tad annoying. But, yeah, hopefully most installs go right the first time!
-
it's a very minor issue. my guess would be some package has more dependencies than it did before, but that's just a wild guess.
-
I didn't install any tool... an interfaces config would have been perfect. But, yeah, pre-generating configs for several tools might be required. Except network-manager. Don't do anything that would encourage anyone to install it. ever. :)
-
I don't know nearly enough about efi to help there. First laptop I've ever owned with it, and I've never set it up... hence why I very quickly noticed the notice didn't have enough info for clueless users like myself. :)
If I have time later I can try another install, but it costs $2 in bandwidth every time. An iso including the base system would make installs easier.
Thanks again!
-
-
mentioned in issue #40 (closed)
-
I believe the fundamental of ascii being uninstallable is now fixed with the latest sysvinit package
@Bushytails : if you feel that the other issues still need addressing, can you please create separate bugs assigned to the relevant projects.
Edited by Daniel Reurich -
Status changed to closed