Dietpi banner upgrades

[timjolson]

Enhancing dietpi-banner.

• Changed custom entry to accept multiple commands
• Added network usage by namespace
• Added disk usage by pattern matching
• Added systemd and fail2ban status
• Standardized printing
• Organized color resources
• Added G_TRUNCATE_MID

[timjolson]

Outstanding items I know of:

• no IPv6 detection for the network traffic, only ipv4
• dynamic disk space usage is independent of the pre-existing disk space entries (what do we do with the existing options?)
• wrapping has not been tested or dealt with
• IP address and network information (retrieval and parsing) use different methods than are pre-existing in the script - need to work through what might be more robust / efficient and consolidate to helper functions
• no handling of permissions checking is in place yet
• there sometimes is an issue with the terminal not being properly cleared when running dietpi-banner 1, probably applies to dietpi-banner 0, as well. The issue occurs often when using VSCode and has not been verified in other terminals.

[MichaIng]

A great extension! But there is some cleanup/polishing needed to reduce the overhead (in code an UI), and make it work for expected use cases.

[MichaIng]

2. dynamic disk space usage is independent of the pre-existing disk space entries (what do we do with the existing options?)

I suggest we merge "Disk usage (RootFS)" and "Disk usage (userdata)" into the new single option. Otherwise it is confusing to have 3 different options all showing some disk usage, and hence also 3 code sections to keep in sync. Maybe / and /mnt/dietpi_userdata (if a mount point) can be dedicated fixed quick-select options in a parent menu. Or, maybe simpler, all findmnt -no SOURCE --real mount points could be part of that menu. So when adding a disk, there is a menu showing all mount points to select + custom, and custom shows the inputbox? The fixed mount points can still be treated like a glob when printing the banner, of course.

[timjolson]

So a dynamic checkbox array that has findmnt -no SOURCE --real options listed, plus a checkbox to then enter custom matching? And the first array item is always /mnt/dietpi_userdata.

[MichaIng]

Right, that would be my suggestion. Little correction: findmnt -nro TARGET --real, to list them by mount point rather than block device, and without the tree view.

The hardcoded /mnt/dietpi_userdata wouldn't even be needed then, as its true mount point would show up in the list already, if it is a dedicated mount. And / shows up at the top, which is good as well.

[MichaIng]

This one seems to be doubled. Especially if the MOTD is disabled, but useful commands or credits are enabled, there will be two lines one after another. If you want to separate the MOTD from the potential prior multi-line outputs, it could be done inside the if (( ${aENABLED[motd]} )) instead.

[MichaIng]

Ah, was this probably meant to be:

(( ${aENABLED[network_usage]} || ${aENABLED[disk_usage]} || ${aENABLED[custom_commands]} )) && echo -e "$CLI_LINE"

So to print a separator line if there was a subheader output above, which do print a separator line before their multi-line output?

[MichaIng]

I do not like that users need to enter all the formatting stuff manually. I suggest the following:

• Turn aCUSTOM_COMMANDS into an associative array, using the subtitle/name as index.
• In the menu, replace "Add/Edit/Remove" with Add/name1/name2/.... So the current custom commands are shown in the menu with their name that is used as array index. Selecting a current command opens the inputbox to edit it. An empty command removes the entry. "Add" requires to enter the command first, and the name afterwards.
• We use Print_Item_State <name> <output>, hence take care of the formatting.

If someone really wants to use a custom formatting, it is still possible to prefix the command with echo -e '\r';, still much shorter than all the color codes needed otherwise.

[timjolson]

My thinking was that a custom command would not have to fit the common formatting, but I am okay with either way. I feel having to name each command is excessive, but that is needed to use Print_Item_State and it's format.

This current version also breaks using custom colors/etc. within the command output, which I think is an important function.
e.g. the alert format for a system condition that is important to the user

I do like an empty inputbox removing the command

[MichaIng]

I feel having to name each command is excessive

Currently, if one wants to have some informational title, it needs to be printed (and in case formatted manually) as part of the command. So the amount of things one needs to type is usually not much different when requiring a name. The other benefit is the better overview in the menu, when active custom commands are visible without entering the "Edit" or "Remove" menu first. Of course numeric names would do as well, like using the current numeric index for the menu entry as well as for Print_Item_State "Command $i" "$output".

This current version also breaks using custom colors/etc. within the command output, which I think is an important function.

The current versions does, indeed. The new version, even with Print_Item_State, would still allow to enter color codes manually, but it wouldn't be needed anymore for the base formatting to match the other banner output lines.

Or we allow both: When leaving the command name/title empty, it is added as numeric index to the array, which triggers the output to be printed directly without Print_Item_State. If it is given a name, Print_Item_State is used for formatting.

[MichaIng]

Just as an overview, the things I'd still want to do before merging. But will do so before DietPi v10.6 release (which will be delayed anyway because of a spontaneous holiday trip consuming some time this week):

• Move array index migration to https://github.com/MichaIng/DietPi/blob/master/.update/patches, so we do not need to carry the code here, but just do this once with a sed loop or so.
• Migrate existing /boot/dietpi/.dietpi-banner_custom and aDESCRIPTION[10] into the new custom command array.
• Migrate/Replace aENABLED[disk_rootfs] and aENABLED[disk_userdata] with aDISK_SPACE_PATTERNS.
• When adding a disk space pattern, findmnt -nro TARGET --real output lines, which are not matched yet, are offered in the menu directly, with "custom" allowing to enter a custom pattern. I'd also list all active pattern in the primary menu, showing the inputbox to edit them on select, removing them by emptying the inputbox, + the "Add" entry.
• Instead of scraping the df output, findmnt -nro TARGET,USED,SIZE,USE% --real might be the cleaner alternative. Though, interestingly it is much slower as I just recognized, taking about 3x more time to process. Alternative is df -h --output=target,used,size,pcent, and we skip the redundancy math + skip separate percent calculation? There is no reason to assume used or size wouldn't be available. Or if used wouldn't be for some special mounts, avail wouldn't be available either. If higher precision and/or aligned unit is wanted the -m flag could be used instead of -h to get MiB values, which can then be converted to GiB slightly simpler.
• Add IPv6 support for network usage and Fail2Ban stats.
• Reduce sudo calls in Print_Network_Usage: If namespaces are used, check via sudo -n -v once whether the current user can use sudo. If so, do the whole namespace processing loop in a single sudo -n dash -c 'declare -f func; func' 2>&1 call. Else, process the default namespace only.

[MichaIng]

I went through the open points you mentioned at the top. Regarding clearing the terminal in VS Code: See G_TERM_CLEAR() in dietpi-globals: It uses ANSI escape sequences, \n and \e[H to move the content into the scrollback buffer (where available) and move the cursor back to the top, rather than clear. I found it preferrable to keep the possibility to scroll back, rather then erasing the content. But it performs worse on certain terminals, where the content is visibly moved line by line. Possible that this does not work well in some dev terminals.