| View previous topic :: View next topic |
| Author |
Message |
pingtoo
Veteran
Joined: 10 Sep 2021
Posts: 1184
Location: Richmond Hill, Canada
|
 Posted: Tue Oct 15, 2024 7:06 pm Post subject: |
 |
|
| Genone wrote: | | One thing to note is that Gentoo profiles cannot be fully stable due to their implementation. Each profile is actually composed of a number of components which are often shared between different "final" profiles, so any change to any parent component will end up in all final profiles using it (unless manually negated). |
I can accept this true, therefor change profile is not a easy thing.
But this just indicate a deficient in design. We should start consider what else can be done without breaking exist practice yet it can bring the to change profile will have some form of notice.
If I can up a idea I will open a thread for discussion. |
|
| Back to top |
|
 |
finoderi
n00b
Joined: 29 Oct 2021
Posts: 62
|
| Posted: Tue Oct 15, 2024 8:50 pm Post subject: |
|
|
| What would be cool to have is a Wiki page that lists all the default USE flags for all the profiles. And ideally that page would update every time there are some changes in profiles, so people who want to know what changed could just check it. |
|
| Back to top |
|
|
GDH-gentoo
Veteran
Joined: 20 Jul 2019
Posts: 1662
Location: South America
|
| Posted: Tue Oct 15, 2024 9:30 pm Post subject: |
|
|
| pingtoo wrote: | | if we want to say the 23 is a year it happen and what is that .0 mean? is it in a indicator it is the first? |
Yep, the second number is also "major" (meaning "user has to perfom a multistep procedure for changing the profile"), and always 0 unless another profile change of that kind happens in the same year.
The only time I saw that happening was 17.0 -> 17.1 (which changed /lib and /usr/lib from symbolic links to real directories). When I installed my first Gentoo the profile was 13.0.
I also don't understand what the complaint is about, no large group of people (such as the universe of Gentoo users) is going to agree on what exactly constitutes a change that should be notified.
Personally, emerge --pretend is all I ever needed, and I don't consider (given the what the current effect is) that gobally enabling the wayland USE flag is a change worthy of a notification, for example...
_________________
| NeddySeagoon wrote: | I'm not a witch, I'm a retired electronics engineer  |
| Ionen wrote: | | As a packager I just don't want things to get messier with weird build systems and multiple toolchains requirements though |
|
|
| Back to top |
|
|
pingtoo
Veteran
Joined: 10 Sep 2021
Posts: 1184
Location: Richmond Hill, Canada
|
| Posted: Tue Oct 15, 2024 9:49 pm Post subject: |
|
|
| GDH-gentoo wrote: | | I also don't understand what the complaint is about, no large group of people (such as the universe of Gentoo users) is going to agree on what exactly constitutes a change that should be notified. |
Please note, it is not a "complaint". I merely observed more than one person raise concern stating that they are not aware of the changes. And to a degree I am with them. so I "suggest" something to be done. And I propose my idea on how to make notice. that is all.
May be it is my poor English skill that lead to seems be a misunderstand of my position. I mean no offence and I am not pushing. I just want to bring out an idea so it can be discussed.
| Quote: | | Personally, emerge --pretend is all I ever needed, and I don't consider (given the what the current effect is) that gobally enabling the wayland USE flag is a change worthy of a notification, for example... |
Yes, that could will be everybody need to do for every time one ran emerge. It just being raise as surprises being the changes were happen at least expected place therefor it become hard to decide what to do. If there were that profile migrate that always have a news item indicate the default state changes, may be there will be lesser people raise questions. |
|
| Back to top |
|
|
pietinger
Moderator
Joined: 17 Oct 2006
Posts: 4960
Location: Bavaria
|
| Posted: Tue Oct 15, 2024 10:01 pm Post subject: |
|
|
pingtoo,
I would like to respond to some of your arguments - that's why I numbered it:
| pingtoo wrote: | [1]
What I can not make peace of it is because consistently saying it is ok to make changes without any indication. this in my opinion is wrong.
[2]
Just because Gentoo is known as rolling release therefor any changes can just happen without notification is wrong.
[3]
What I have been propose is using the profile naming convention to indicate changes happen therefor one should pay attention as well gave someone a opportunity to choose to opt in.
[4]
[...] And this is where I am trying to convince that the changes in profile is bigger enough that deserve a notification. I am just keep on saying the it is easier just use the profile naming convention to indicate the changes.
[5]
However if the default changes from time to time without a easy indicator then it will require support as for further question that when did you do this before a known "changes" or after the known "changes". |
A) But first i would like to draw your attention to this thread: https://forums.gentoo.org/viewtopic-t-1171174-highlight-.html
The user gave us the output of “emerge -vuDU @world” AND “emerge --info” in his initial post.
There you saw that he uses ACCEPT_KEYWORDS=“amd64 ~amd64”.
In the end, the cause of his problem was this: /etc/portage/package.accept_keywords:dev-qt/qtwebengine -~amd64
Found by @bstaletic - I was too blind ... BUT ... I requested this:
| Code: | | # grep -r --exclude-dir=savedconfig --exclude-dir=repos.conf "" /etc/portage |
So @bstaletic was able to find above reason.
For many, many years (I can't remember since when) it has been possible to make settings not only globally, but also specifically on a per-package basis (you know that too) ... and not only for the use flags, but also other things like the accepted keywords or various environments. And as you and I know, there is probably no Gentoo user who does not have at least one change in make.conf for his use-flags ... and not only globally in make.conf but also specifically for some packages.
B) Before the “wayland” use flag was activated globally (for desktop profiles), it was already in use: It was specifically activated (by the profile) for the “mesa” package. I noticed this at the time by checking what WOULD happen (with parameter -p) every time BEFORE I perform a world update. I can see every change to use flags immediately, as they are displayed in green or yellow. If I then check with “emerge --info” whether it is activated globally, and this is not the case, then I know that it was activated specifically for this package by the profile. Should this change be indicated by a version change of the profile? I suspect not (I also suspect it would not be technically feasible to map every change of use-flags to version numbers). Some time later the use-flag “wayland” was activated globally and you could see it in “emerge --info”. How big is the difference compared to before?
Now I would like to answer you:
1. I think there is ALWAYS a notification! It's just a question of WHERE. And I say, if you do an “emerge -uUDvp @world” before the world-update itself is done, you will immediately see that something has changed in the use-flags.
2. I completely agree with you ... but that is exactly not the case.
3. We both know that there are always users who do not do an “eselect news read”, even if portage points this out. And I think that this is an even clearer indication that something has changed than a “small” change in the profile version numbers.
4. I cannot judge how much effort is involved in mapping every change of use flags in the profiles, but I assume that it is immense. (*) Just remember that the use-flag “vulkan” was recently activated globally (next to “qt6”). However, I always noticed this in good time, as it is visible in “emerge -uUDvp @world”. There was even NEWS about the use-flag “initramfs” ...
5. Here I completely disagree: For most problems, “emerge --info” alone is NOT enough. My example from (A) is actually the normal case and not an exception ... for support we always have to check the specific configuration ... and we need more than just a profile number...
*) pingtoo, I see your good intentions, but I don't think the cost-benefit ratio is good here.
_________________
https://wiki.gentoo.org/wiki/User:Pietinger |
|
| Back to top |
|
|
pingtoo
Veteran
Joined: 10 Sep 2021
Posts: 1184
Location: Richmond Hill, Canada
|
| Posted: Tue Oct 15, 2024 11:40 pm Post subject: |
|
|
pietinger,
First, thank you very much for your direct response. I really appreciate those points.
Total understand and agree for that thread that the original poster had misunderstand the condition, therefor he/she draw wrong conclusion. I think however it is worthy of discussion on why they draw the wrong conclusion (especially compare to if the default did not changed)
So forgive me if I interpreted you argument in wrong way.
My sense of you argument is that current setting is good enough and change to profile is harmless therefor it should be freely do so.
This is where we see it different, I am seeing as it "could" be better. I am saying if at the point of profile naming made it different (I accept that we will not use the word "version). It will make a lot easier to notice. And furthermore it even give chance to opt in.
Yes, many don't read news item, (Part of reason I suggest use profile naming change). But I think you arguments have missing the fact that if the profile default did not changed but introduce as a new profile than the reaction will be quite different. Because if someone said "after switch to the new profile" following happen that stop working and they think it should work. Then the argument that Gentoo is rolling release and it is evolving make a lot more stronger argument. (Because their choose to use new profile that have developer intent for better default.)
I don't know the effort of making profile change. So I just want to use the idea to introduce to said change default without notification is wrong. I understand that one just running emerge -uUDvp @world should/would let one know about the changes. But have you thought about that if the profile default did not changed what is that view would be? And just that there is a now profile that should one choose to move on to will be so much easier for everybody. (as those choose to stay with old default)
Any one use Gentoo for sometime already learned from running emerge -uUDvp @world when ebuild got new release for what they chooses in their package set. However It is usually mystery for someone to notice that the normal few packages change update all of certain become a very large list of thing to do.
I am NOT saying from this point onward developer *should* not make change to profile any more. I just want to brought to attention so when the decision to make change profile should be consider more carefully before proceed.
I thank you for give me the oppotunity to better explain my thoughts. |
|
| Back to top |
|
|
pietinger
Moderator
Joined: 17 Oct 2006
Posts: 4960
Location: Bavaria
|
| Posted: Tue Oct 15, 2024 11:58 pm Post subject: |
|
|
| pingtoo wrote: | | I thank you for give me the oppotunity to better explain my thoughts. |
I thank you just as much for your explanations.
_________________
https://wiki.gentoo.org/wiki/User:Pietinger |
|
| Back to top |
|
|
finoderi
n00b
Joined: 29 Oct 2021
Posts: 62
|
| Posted: Wed Oct 16, 2024 5:10 am Post subject: |
|
|
| That person did outright stupid shit like globally setting both ~amd64 and amd64 and -qt5 and -qt6 and no amount of warnings in profile would save him from him. |
|
| Back to top |
|
|
Zucca
Moderator
Joined: 14 Jun 2007
Posts: 3652
Location: Rasi, Finland
|
| Posted: Wed Oct 16, 2024 6:40 am Post subject: |
|
|
@finoderi: Your post on the other hand does not add anything to this topic. Watch your language and be respectful.
_________________
..: Zucca :..
Gentoo IRC channels reside on Libera.Chat.
--
| Quote: | | I am NaN! I am a man! |
|
|
| Back to top |
|
|
tld
Veteran
Joined: 09 Dec 2003
Posts: 1845
|
| Posted: Wed Oct 16, 2024 12:25 pm Post subject: |
|
|
I hadn't had a chance to reply to my own thread...sort of took on a life of it's own.
I'm not sure I have any specific opinion as to the necessity of global USE chnages to existing profile versions. What did strike me though was that it's happened twice to me recently on x86 (once with qt6 and once with wayland), and in over 20 years using Gentoo I'm not sure I recall it happening ever before. That seemed odd.
Tom |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2317
|
| Posted: Wed Oct 16, 2024 12:43 pm Post subject: |
|
|
| tld wrote: | and in over 20 years using Gentoo I'm not sure I recall it happening ever before. That seemed odd.
|
It seemed odd because it happened a lot. It just didn't happen to be something you care about and that's why you don't remember it. It's because you didn't notice it.
It happens quite a lot and quite regularly. Most obvious examples being python and ruby targets, masked/unmasked flags and so on. All those happen in the profile.
TBH after the first few comments from pingtoo I started thinking the same way as you and him but then I helped somebody to migrate to ruby 3.2 and it occured to me all those things were happening in the profile, we just didn't realize it. In fact a lot of what you're dealing every day with happens in the profile.
Best Regards,
Georgi |
|
| Back to top |
|
|
finoderi
n00b
Joined: 29 Oct 2021
Posts: 62
|
| Posted: Wed Oct 16, 2024 3:27 pm Post subject: |
|
|
| Zucca wrote: | | Your post on the other hand does not add anything to this topic. |
That's because you decided to read it that way. That link itself is a rather poor example and has little to do with profiles.
But lack of knowledge what flags are enabled by default in a profile you are using is a real problem. And there is no clear way to check it other than $emerge --info, which doesn't differentiate between default and deliberately set flags. |
|
| Back to top |
|
|
tld
Veteran
Joined: 09 Dec 2003
Posts: 1845
|
| Posted: Wed Oct 16, 2024 3:53 pm Post subject: |
|
|
| logrusx wrote: | | It happens quite a lot and quite regularly. Most obvious examples being python and ruby targets, masked/unmasked flags and so on. All those happen in the profile. |
Well yea, for sure I've seen and expected those forever. That's VERY different from qt6 and wayland which at least at this point very much ARE user preferences. That's just plain annoying.
Tom |
|
| Back to top |
|
|
asturm
Developer
Joined: 05 Apr 2007
Posts: 9230
|
| Posted: Wed Oct 16, 2024 4:02 pm Post subject: |
|
|
| It does not matter if they are not *your* preferences. They are the preferences that make the most sense for a desktop profile governing the package repository at this point in time. |
|
| Back to top |
|
|
pingtoo
Veteran
Joined: 10 Sep 2021
Posts: 1184
Location: Richmond Hill, Canada
|
| Posted: Wed Oct 16, 2024 4:18 pm Post subject: |
|
|
| asturm wrote: | | It does not matter if they are not *your* preferences. They are the preferences that make the most sense for a desktop profile governing the package repository at this point in time. |
I am not sure I understand the point this statement trying to make. It is very different from the objection other bring out.
There is nobody objecting making those changes.
The objection is the act of change without notice surprise people. basically those bring out the objection stated that they wasn't' expect this happen when they did not choose so. They are not objecting that having qt6/wayland is wrong.
| Quote: | | They are the preferences that make the most sense for a desktop profile governing the package repository at this point in time |
If the preference at this point of time is with those changes, should we not have a marker to state that previous preference is that (without qt6/wayland)? |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2317
|
| Posted: Wed Oct 16, 2024 4:22 pm Post subject: |
|
|
| tld wrote: | | That's just plain annoying. |
Do you realize you're not the only Gentoo user and your preferences are not the only user preferences out there?
I remember back when I installed the installation that served for a basis of my current one, qt5 was default enabled. I just disabled it in make.conf as a result of inspecting what was about to be emerged and researching what pulled it in.
Particularly the case with qt6 is because conflicts should be resolved one way or another. Would you be happy if it was qt5 default enabled? But now that qt6 is stable it wouldn't be a good idea.
Wayland is harmless. It's been discussed numerous times what the results from enabling it are.
Best Regards,
Georgi |
|
| Back to top |
|
|
asturm
Developer
Joined: 05 Apr 2007
Posts: 9230
|
| Posted: Wed Oct 16, 2024 4:22 pm Post subject: |
|
|
| pingtoo wrote: | | I am not sure I understand the point this statement trying to make. |
You know what a direct response to the previous post is if you do not see a quote. You do not have to follow up *every* *single* utterance with your repetitive misunderstandings of what profile upgrades do and can't do.
Last edited by asturm on Wed Oct 16, 2024 4:24 pm; edited 1 time in total |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2317
|
| Posted: Wed Oct 16, 2024 4:24 pm Post subject: |
|
|
| pingtoo wrote: |
The objection is the act of change without notice surprise people. |
No, that's not the objection. The objection is against the change itself. Anyway, things in the profile change all the time and notices would be mundane to have for everything.
Best Regards,
Georgi |
|
| Back to top |
|
|
asturm
Developer
Joined: 05 Apr 2007
Posts: 9230
|
| Posted: Wed Oct 16, 2024 4:51 pm Post subject: |
|
|
| logrusx wrote: | | Would you be happy if it was qt5 default enabled? But now that qt6 is stable it wouldn't be a good idea. |
If you really want to know, both are currently enabled, and will remain so for a long time. It is the responsibility of ebuild maintainers to avoid conflicts as much as possible, and where conflicts exist, desktop profile defuses them with a specific package.use entry.
Qt porting is not instantaneous, it cannot be. So, you're starting from a point where
- 100% of Qt-based packages are Qt5-based, regardless if they even have USE=qt5, USE=gui or none of them
- Then, libraries are being ported to Qt6, along with applications, and desktop environments, ideally remaining in ~arch
- Conflicts! So many conflicts to figure out, because upstream developers often can not (since they happen down their dependency chain they have no direct control over) or do not think about a distribution that provides both major versions of desktops/applications/libraries in the same package tree in different speeds (arch, ~arch) for users to potentially mix and match and especially lets them build from source
- From that moment on, desktop maintainers need to consider that more and more applications are starting to look out of place in their Qt5-based desktop environments, because they cannot properly theme those new Qt6 arrivals, but their shiny new Qt6-based desktop may not be ready yet
- ~arches need re-keywording for all the dependencies that have changed, moved, were added
- Maybe at some point you reach 50:50 between applications that are Qt5-based and those that were ported to Qt6 in ~arch
- That is the time for desktop environment maintainers to think of the tipping point where they need to have their products ready (and hopefully upstream are too), and a big stabilisation push is initiated
- New things are stabilised, Qt5-based packages and with them blockers that needed to be raised against conflicts can be cleaned up again, so bug noise caused by arch/~arch mixing users as well as legitimate bug count is getting lower again
- So then Qt6-based desktop environments are stable, and Qt6 integration in other desktop environments is stable, so surely everyone is happy!
- No! You still need Qt5, since not everything has been ported yet, and they (the libraries) work just fine side by side anyway, and having a hard cut-off between Qt5 and Qt6 would simply rob users of GUIs and compatibility, compared to our starting point
- ...until after a very long time the totality of Qt packages will be Qt6-based,
- and Qt5 can be cleaned up from ::gentoo.
And that's just one toolkit upgrade, and an extremely simplified representation of events. All of these events are gradual, rolling, partially overlapping, up to the assessment of maintainers, need careful preparation between ~arch and arch, timing with upstreams and other package maintainers, ...
And that's what I mean with you cannot disconnect profile defaults from ebuilds at a given point in time ... |
|
| Back to top |
|
|
pietinger
Moderator
Joined: 17 Oct 2006
Posts: 4960
Location: Bavaria
|
| Posted: Wed Oct 16, 2024 5:40 pm Post subject: |
|
|
asturm, thank you very much for your description which gives a good insight into the work of you (and all other Gentoo developers). Posts like this are extremely helpful in promoting understanding among our users.
I can still remember the changeover from kde 3 to 4 and then to 5, and have the feeling that the current changeover to 6 will be the least painful. Maybe even the upgrade that won't cause any problems at all (as long as users don't try to be smarter than our developers).
_________________
https://wiki.gentoo.org/wiki/User:Pietinger |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2317
|
| Posted: Wed Oct 16, 2024 5:45 pm Post subject: |
|
|
| pietinger wrote: | | Posts like this are extremely helpful in promoting understanding among our users. |
Yes! Thank you, astrum for the detailed explanation. So much work happens behind the scenes users are unaware and take for a given.
Best Regards,
Georgi |
|
| Back to top |
|
|
|
Portage & Programming
