Let's continue to contribute and report bugs where it's appropriate.

@ eric I start by linking to the requested .dialogrc I created exclusively for this. It's organized in meaningful blocks and well commented for anybody interested in what variables do what. Some are pretty weirdly pulled by several dialog boxes in different manners.

And a question: I suppose it is not easy/possible to change the color default variables of the console (BLACK,WHITE,RED etc.) to non-default values of some sort outside of an X server? Because look at the difference a little bit of subtle coloring does:

Seriously now, that part of the installer where it creates the /fstab for the installation, should it have the -U option on genfstab so it makes it with UUIDs instead of /dev/sdxx ??

The 1st screen about kbd selection is now gone, right?
I did a base install pretty default stuff.
At some point the installation hang at this screen


When I hit return it was asking for the second entry of the root pw, so it didn't match and it was redone .. seems like a prompt/return to installer screen missing step.
Everything else went smooth. Should the advanced options (editing /etc/66/boot.conf ...) be enforced at the end of the installation or remain optional?
Please update it and play with it in particular the auto-part menu
it looks like this

if you unselect all the result will be a simple root partition
if you only select boot the result will be a /boot /root partition
if you only select swap -> /root /swap
only home -> /root /home
if you select boot swap the result will be /boot /swap /root
and so on...
All combinaison is possible, please play with this bad boy and make reports

Note: if you come from UEFI the boot options is gone. The boot partition will be mandatory

For the moment i concentrate my effords about the code not the design but i think that it's difficult to see the difference between white and bold white mostly for the <select><back><done>... bouton
i don't know if the choice of replacing the number (on a menu) by the first word of the sentence is a good idea or not. The choice should be picked between : one sentence or one word and sentence e.g:
"disk(s)" "prepartion (optional)"
or
"disk preparation (optional)"

@ At some point the installation hang at this screen
should be fixed, tell me it

@ Should the advanced options (editing /etc/66/boot.conf ...) be enforced at the end of the installation or remain optional?
the essential configuration part is made automatically on the file by the customize chroot menu (when you define hostname,keymap,...). So a new user should not care about this, an advanced user know the options

@ The 1st screen about kbd selection is now gone, right?
you can activate it again by the /etc/obarun/install.conf file with the KEYBOARDCHECK variable if you want
eric wrote[...]it's difficult to see the difference between white and bold white[...]
[...]i don't know if the choice of replacing the number (on a menu) by the first word of the sentence is a good idea or not[...]
# 1 this is an issue which I have seen in X as well and does not happen in tty console since the default non-bold white is a medium grey there. A workaround could be to use bold black for inactive content which also is a darker grey. Must be evaluated carefully though. If you wanted to test this real quick it would suffice to set screen_color to (BLACK,BLACK,ON) instead. Don't be scared by the looks of the variable and see for yourself :)
Telling from the background color and the pixel rendering of that screenshot I assume you are showing this from inside live JWM? I could make a color profile for you to include into the live JWM terminal to resemble the upper most screenshot of post # 1 which has very nice contrast.

While testing I see that you're quite right with the contrast even in tty, but the real culprit here is the separation of one single line in <tag> and <item> and only <tag> being highlighted. I suppose you did this to preserve the [variable] color of green for it would become white while being selected. Would that be such a big issue? I'll look further into this.

# 2 not a good idea. Just use --no-tags and make the whole sentence be a single <item> attribute or you'll heavily interrupt legibility. In you screenshot only the <tag> element is bold, why? The .dialogrc should actually make the selected <item> element bold as well to begin with.

Oops found it. Please change .dialogrc line 90 from:
item_selected_color = screen_color
to
item_selected_color = button_active_color

Everything else: I'll make a test run later, sounds great. Fungal will probably be faster ;P

With the given (and corrected [line 90...]) .dialogrc its just a matter of playing around with three variables (TRON anybody? :D):

screen_color = (WHITE,BLACK,ON) <- will be all default text not selected including messages and inner borders
border_color = (WHITE,BLACK,ON) <- will be the outer border of course
button_active_color = (CYAN,BLACK,ON) <- dictates all selected elements

With this theme just make sure that the contrast between screen_color (inactive) and button_active_color (active) is good enough as you mentioned and try not to have the eye get focus on the border too much.


Functionality observations
  • Cancel buttons appear to always throw the user back to mainMenu() - would be better UX to go the menu from before
  • If I choose to have a separate swap partition I can enter the whole disk size and still get to the next step which should not be the case.
    You may argue that one can not be stupid enough to do this in a real situation hahaha,...maybe could do a test for if rest is >= 5G for root to be possible to install, or just let it be :D (probably the better thing to do for now)
  • I chose XFS for root and BTRFS for /home (yeah I know...) and the script bailed out with:
    parted: invalid token: vfat
    Error: Expecting a file system type.
    (Now tested with various, seems to not be related to a specific FS, probably a spelling mistake in the parted command?)
    Further playing shows that this is inside of some UEFI parted command only. The BIOS path seems to partition fine
So, the code itself is almost done (nothing new shared from my last post for you). Let's talk about design.
We made the choice to remove number.Well, now we have two solutions: tags or not. I prefer the version without tags but this hit the limit of dialog possibilities.
If i don't use tags i can't use color to specified [default values] into the message because dialog consider it as the tag itself instead of a message. The resulting message will content the escape character instead of the colors like e.g this \Z\n. This point is clearly annoying for me because visualizing speaking the eyes is automatically attracted by the color.
Also, without tags i can't pass to the next menu options automatically. I mean if you enter on the e.g disk(s) partition menu with you exit from it and come back to the main menu the next entry is automatically took. This is not an important for us but for user i think it's a good point and avoids to go down from the start of the menu list every time.
Now, i think i'm good with the color scheme, isn't it?(see below)



Seriously now, that part of the installer where it creates the /fstab for the installation, should it have the -U option on genfstab so it makes it with UUIDs instead of /dev/sdxx ??
i will change this behavior

A word about the fact to restart the installer after an update:
For the moment i don't find a clean solution (and safe). I prefer to annoy one time the user instead to have a crash due of a out of date function
  • I think it's completely legit to have to annoy us if there's no elegant solution to the update situation :)
  • Scheme is all well if you're ok with it that way
  • The escaping situation was actually something I too stumbled upon and I was like "how did Eric did that then?" Oh well now I know...
    So I guess there's no choice but numbering the entries because:
    • A phrase can not be spread apart like that in most textual context, even less so in a book or a user interface
    • Having show the escaping sequence is apparently a no-go
    • The only alternative would be to not color the default values which is a main feature of your design idea
The downside I see to numbers is the very prominent implication that you have to take steps 1-x to complete anything when in reality that is not the case. I guess that's dialog then. I think it's coming out great.

Edit:
According to this post the root entry in fstab is now mounted "rw" even by the installer? Didn't you say once that we should mount "ro"?
Please guys update and torture it, it's ok for me now.Obviously i expect some change about syntax, grammar and sentence ;)
Note: bootloader and fstab use now UUID

@ Didn't you say once that we should mount "ro"?
Yes, i found the bugs, you reverse the sed replacement ;), i fixed it now

Please, try to deal with UEFI
eric wrote @ Didn't you say once that we should mount "ro"?
Yes, i found the bugs, you reverse the sed replacement ;), i fixed it now
So now it should be rw in the boot command?
So now it should be rw in the boot command?
no, ro was always the choise by default and it will be kept as it. So / is mounted ro ALWAYS, then at some part the boot service set will mount it rw
Please udpate again, bootloader installation was bugged, should be fixed now
I always had to switch manually from rw to ro, it was always done by default as rw. Am I wrong?
Was it always bugged, or are we talking about a different thing? I am speaking of grub.
Ok, grub comes out as ro now.
1 For each of "flavors" of installation I recommend the minimum disk space is written next to that label (openbox 3.5GB), minimal (1.6GB), plasma (45GB :), JWM (3.56GB) .. xfce4 (4.2GB). Or the installer should check on the partition sizes and say in advance whether it is going to be enough, not begin the installation and then half way through find out it is running out of space.
2 The lists of software: Either make one big one or something different. If I keep running into errors after 2y I can imagine any new user trying it for the first time will at least have just as much. I think it is a bad idea to list one piece of software AND its dependencies, it makes lists long. For example, openntp-66serv doesn't need openntpd, or connmand-66serv doesn't need connman. It is very easy to get into editing these lists and add something that creates a conflict, then the installer crashes, and I don't think it provides sufficient information of what really caused the crash (specific conflicts).
3 I decided to try this on a vm and a clean disk. I started the installer once and chose the assisted (expected auto partioning - not the spanish inquisition as the Montys would say). Bloop! Ok, I went and made a GPT partition table. Bloop, Back out I went and created boot, swap, root, home. Bloop, it tried to recreate them, bloop, bloop. The hell with autopartition. So I tried the manual definition of partition and it worked out. I made my first bootable -efi installation. Openbox failed the first time, I undershot the required space by 5% so it failed the 2nd stage after it then figured out space was insufficient, when base was installed already. Bloop! Switched to minimal and no more bloops.

4 What is it with glibc upgrading? Is that absolutely necessary? Could this have affected something on how jwm was running? Something collapsed during installation, gparted and other things were reporting "failure to connect". All services were running fine, I exited and restarted jwm and it was ok again. This was a first ever such behavior on liveJWM.

I forgot to record it somewhere, when the installer goes out on terminal/console, after the base files have all been installed and it is trying to install the rest of the system, X and other tools, it says check base installed OK, it should then say checking of packages to be installed not checking installed, which are not installed yet. It may be a little confusing to someone on what the installer is doing.

EDIT: I tried booting the live in vm as efi and all options failed for me. On the same setups when I did the efi installation it booted the real installation, it was only the live that failed (JWM 102019). I've never done such a thing before, hating efi and all, so I don't know whether this is known or it is a strange occurrence.
Uff Fungal, that's quite interesting a read ;)
I like the idea of # 1 except for your estimates :D
I'll test later though, pretty tired again...

Btw. I think the assisted method is a good idea (that apparently needs fixing), but I suppose if anybody could write a comprehensive auto-partition function that respects the rest of the code it would be welcomed by Eric to be included in the future ;)
@ 1
added, a check of the size depending of the template is made at every size entry choice

@ 3
are you sure to use the last commit. Anyway, found some trouble, update it , it seems to be ok now

@ 4
The trouble comes at password entry. If the glibc into the ISO is older than the version installed on the fresh installation, the function enter in a loop (password to not match). So to be sure that this kind of stuff do not happen the glibc is automatically updated if you come from an ISO

@ 2
Yes, it a little confusing for new (or old user :p). I can't works on it for the moment. i will make change at the next release. However, this is an optional step and new user should not touch it.
I forgot to record it somewhere, when the installer goes out on terminal/console, after the base files have all been installed and it is trying to install the rest of the system, X and other tools, it says check base installed OK, it should then say checking of packages to be installed not checking installed, which are not installed yet. It may be a little confusing to someone on what the installer is doing.
Well, i can't make this distinction because, this check is always made even if the base or not are installed. This check was created to be able to install again from the last crash when a previous installation fail. So, i can't know if the last previous try has failed or not. Hope i'm clear.
eric wroteWell, i can't make this distinction because, this check is always made even if the base or not are installed. This check was created to be able to install again from the last crash when a previous installation fail. So, i can't know if the last previous try has failed or not. Hope i'm clear.
Ok, so this comes out on the screen but if everything proceeds ok it shouldn't matter, but if something fails some of the X parts may be installed some not, and it checks what hasn't been installed. Got it, I think.

I think it is ready to fly, I even like the multicolored menu as it is now.
the bad boy was merged to dev branch, if you want to follow and play with it please update from this branch instead of dev-eric, thanks
mmmm.... I don't like it, maybe with specific colored bg of terminals it is even worse, but I'd rather have the multicolored or pretty much any thing pictured on this thread. The tint difference between red and bright/bold red is very little on this screen. If working on the bridge at night on an ocean liner this may be good :)

Powered by Obarun