| View previous topic :: View next topic |
| Author |
Message |
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
 Posted: Mon Sep 04, 2006 11:02 pm Post subject: |
 |
|
| Gergan Penkov wrote: | | well peace, I would be happy to test how paludis reverse-dependancies work, when they are implemented. |
It's dead easy. It checks system, and RDEPEND and PDEPEND of things that're installed, and doesn't try guessing based upon dodgy hacks like NEEDED that'll miss zillions of dependencies. |
|
| Back to top |
|
 |
Gergan Penkov
Veteran
Joined: 17 Jul 2004
Posts: 1464
Location: das kleinste Kuhdorf Deutschlands :)
|
| Posted: Mon Sep 04, 2006 11:19 pm Post subject: |
|
|
Well thought not responding, but emerge gnome with howl use-flag (at least it was so named some 6 months ag), recompile all gnome packages 2-3 times, remove howl and the use-flag (no matter with/without reverse dependancies implemented), reemerge with --newuse or whatever - you have a broken gnome installation, with consequences similiar to the expat one.
This kinda breaks the idea of proxying things - you have two-three libraries, which interact with howl not all of them. And yep this breaks revdep-rebuild as it knows nothing about these dependancies - so we are back where we started - the dependancies, in the ebuild and in the configure, probably are mostly correct on first install, but not on subsequent recompilations - this is the problem of not imposing --as-needed. And all was relatively good so long there were only binary distributions and that much possible dependacies, but now for example fedora and debian have begun to also use --as-needed...
_________________
"I knew when an angel whispered into my ear,
You gotta get him away, yeah
Hey little bitch!
Be glad you finally walked away or you may have not lived another day."
Godsmack |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Mon Sep 04, 2006 11:23 pm Post subject: |
|
|
| Gergan Penkov wrote: | Well thought not responding, but emerge gnome with howl use-flag (at least it was so named some 6 months ag), recompile all gnome packages 2-3 times, remove howl and the use-flag (no matter with/without reverse dependancies implemented), reemerge with --newuse or whatever - you have a broken gnome installation, with consequences similiar to the expat one.
This kinda breaks the idea of proxying things - you have two-three libraries, which interact with howl not all of them. And yep this breaks revdep-rebuild as it knows nothing about these dependancies - so we are back where we started - the dependancies, in the ebuild and in the configure, probably are mostly correct on first install, but not on subsequent recompilations - this is the problem of not imposing --as-needed. And all was relatively good so long there were only binary distributions and that much possible dependacies, but now for example fedora and debian have begun to also use --as-needed... |
Eh? Your argument is along the lines of "we should paint the roads pink because terrorism is the number one killer of gerbils, and also Switzerland is doing it". You're talking about lots of different things that aren't really connected. |
|
| Back to top |
|
|
Gergan Penkov
Veteran
Joined: 17 Jul 2004
Posts: 1464
Location: das kleinste Kuhdorf Deutschlands :)
|
| Posted: Mon Sep 04, 2006 11:29 pm Post subject: |
|
|
But show at the end how usefull --as-needed is, if you have a better solution to these problems - please present them 
And how much flawed the logic is to leave everything to the developer...
_________________
"I knew when an angel whispered into my ear,
You gotta get him away, yeah
Hey little bitch!
Be glad you finally walked away or you may have not lived another day."
Godsmack |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Mon Sep 04, 2006 11:33 pm Post subject: |
|
|
| Gergan Penkov wrote: | | But show at the end how usefull --as-needed is |
What? No it doesn't. All your argument demonstrates is that, as expected, most people using --as-needed don't have a frickin' clue what it does, how it works, what problems it solves, what problems it creates, how Unix linking works, what ELF NEEDED is, what types of non-ELF dependencies exist, how dlopen() works, what an ABI is, why upgrading expat broke their systems or why you shouldn't believe everything you read in documentation that was written by people who don't know half of those things either.
| Quote: | | if you have a better solution to these problems - please present them |
Leaving everything to the developer, plus slots and proper QA. |
|
| Back to top |
|
|
Gergan Penkov
Veteran
Joined: 17 Jul 2004
Posts: 1464
Location: das kleinste Kuhdorf Deutschlands :)
|
| Posted: Mon Sep 04, 2006 11:40 pm Post subject: |
|
|
| ciaranm wrote: | | Gergan Penkov wrote: | | But show at the end how usefull --as-needed is |
What? No it doesn't. All your argument demonstrates is that, as expected, most people using --as-needed don't have a frickin' clue what it does, how it works, what problems it solves, what problems it creates, how Unix linking works, what ELF NEEDED is, what types of non-ELF dependencies exist, how dlopen() works, what an ABI is, why upgrading expat broke their systems or why you shouldn't believe everything you read in documentation that was written by people who don't know half of those things either.
| Quote: | | if you have a better solution to these problems - please present them |
Leaving everything to the developer, plus slots and proper QA. |
Thanks god, it was good that I haven't heard of your arguments before, because I would have then the beatiful possibility to go through the expat hell, and no slotting does not have anything to do with this, it does not resolve the problem it only circumvent it. And yes I know most of the things, which you have mentioned.
EDIT: hehe someone fixed it, thanks guys
_________________
"I knew when an angel whispered into my ear,
You gotta get him away, yeah
Hey little bitch!
Be glad you finally walked away or you may have not lived another day."
Godsmack
Last edited by Gergan Penkov on Fri Sep 08, 2006 11:00 pm; edited 1 time in total |
|
| Back to top |
|
|
dleverton
Guru
Joined: 28 Aug 2006
Posts: 517
|
| Posted: Tue Sep 05, 2006 10:33 am Post subject: |
|
|
| ciaranm wrote: | | One example? When a binary dlopen()s a library in such a way that that library requires a third library that as-needed has removed from the needed list. |
So why not link the dlopen()ed library against the library it requires? |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Tue Sep 05, 2006 11:09 am Post subject: |
|
|
| dleverton wrote: | | ciaranm wrote: | | One example? When a binary dlopen()s a library in such a way that that library requires a third library that as-needed has removed from the needed list. |
So why not link the dlopen()ed library against the library it requires? |
Uh, doesn't make any difference if as-needed makes the linker remove it. |
|
| Back to top |
|
|
dleverton
Guru
Joined: 28 Aug 2006
Posts: 517
|
| Posted: Tue Sep 05, 2006 1:27 pm Post subject: |
|
|
| ciaranm wrote: | | dleverton wrote: | | ciaranm wrote: | | One example? When a binary dlopen()s a library in such a way that that library requires a third library that as-needed has removed from the needed list. |
So why not link the dlopen()ed library against the library it requires? |
Uh, doesn't make any difference if as-needed makes the linker remove it. |
OK, maybe I'm misunderstanding the docs for --as-needed, or the situation you described. I thought you meant that binary A links against libB, but doesn't use any of its functions directly. At runtime, it loads Cplugin, which does use libB. If A was linked with --as-needed, the linker will ignore libB, then Cplugin will fail to load, right? But if Cplugin is linked against libB, won't the linker see that it's needed, and arrange to load libB at the same time as Cplugin, no matter what options are used? |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Tue Sep 05, 2006 1:34 pm Post subject: |
|
|
| dleverton wrote: | | OK, maybe I'm misunderstanding the docs for --as-needed, or the situation you described. I thought you meant that binary A links against libB, but doesn't use any of its functions directly. At runtime, it loads Cplugin, which does use libB. If A was linked with --as-needed, the linker will ignore libB, then Cplugin will fail to load, right? But if Cplugin is linked against libB, won't the linker see that it's needed, and arrange to load libB at the same time as Cplugin, no matter what options are used? |
Except if Cplugin's link to libB is removed by as-needed, which can happen in cases where Cplugin's use of libB is only triggered by it being loaded by A, and where A does not use libB directly itself. |
|
| Back to top |
|
|
Gergan Penkov
Veteran
Joined: 17 Jul 2004
Posts: 1464
Location: das kleinste Kuhdorf Deutschlands :)
|
| Posted: Tue Sep 05, 2006 6:45 pm Post subject: |
|
|
From what I gather from these examples, are you trying to implement interposing in this rather hackish way?
_________________
"I knew when an angel whispered into my ear,
You gotta get him away, yeah
Hey little bitch!
Be glad you finally walked away or you may have not lived another day."
Godsmack |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Tue Sep 05, 2006 7:08 pm Post subject: |
|
|
| Gergan Penkov wrote: | | From what I gather from these examples, are you trying to implement interposing in this rather hackish way? |
No. |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 2632
Location: Bay Area, CA
|
| Posted: Sun Sep 10, 2006 6:03 am Post subject: |
|
|
| does paludis do parallel merges? |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Sun Sep 10, 2006 10:39 am Post subject: |
|
|
| devsk wrote: | | does paludis do parallel merges? |
No. Not at the moment, anyway. |
|
| Back to top |
|
|
cokehabit
Advocate
Joined: 23 Apr 2004
Posts: 3302
|
| Posted: Mon Sep 11, 2006 12:29 pm Post subject: |
|
|
paludis --isnt needed  |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Mon Sep 11, 2006 2:39 pm Post subject: |
|
|
| cokehabit wrote: | | paludis --isnt needed |
Ah, that's where you're so very wrong. Who else is going to come up with a solution to the "Portage doesn't do all the things developers need" problem? |
|
| Back to top |
|
|
cokehabit
Advocate
Joined: 23 Apr 2004
Posts: 3302
|
| Posted: Mon Sep 11, 2006 3:11 pm Post subject: |
|
|
| ciaranm wrote: | | cokehabit wrote: | | paludis --isnt needed |
Ah, that's where you're so very wrong. Who else is going to come up with a solution to the "Portage doesn't do all the things developers need" problem? |
so paludis is for developers then |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Mon Sep 11, 2006 3:23 pm Post subject: |
|
|
| cokehabit wrote: | | ciaranm wrote: | | cokehabit wrote: | | paludis --isnt needed |
Ah, that's where you're so very wrong. Who else is going to come up with a solution to the "Portage doesn't do all the things developers need" problem? |
so paludis is for developers then |
Not at all. Paludis provides the things that developers need to give users a decent experience. Obvious examples include [use] and :slot deps, which have little (but not quite no, for :slot) direct value to the end user, but will eliminate all kinds of annoying compile errors and massively reduce the number of times the user sees a hugeass die in the middle of a long series of package builds. Another is the package use masking and both kinds of use forcing -- again, no direct use to end users, but they eliminate cases where a user installs a package with USE="foo" and then has to wonder why foo support is turned off. |
|
| Back to top |
|
|
ferringb
Developer
Joined: 03 Apr 2003
Posts: 355
Location: USA
|
| Posted: Wed Sep 13, 2006 8:30 pm Post subject: |
|
|
| ciaranm wrote: | | cokehabit wrote: | | ciaranm wrote: | | cokehabit wrote: | | paludis --isnt needed |
Ah, that's where you're so very wrong. Who else is going to come up with a solution to the "Portage doesn't do all the things developers need" problem? |
so paludis is for developers then |
Another is the package use masking and both kinds of use forcing |
Worth noting portage supports this already...
_________________
I don't want to be buried in a pet cemetery. ~Ramones |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Wed Sep 13, 2006 8:37 pm Post subject: |
|
|
| ferringb wrote: | | ciaranm wrote: | | cokehabit wrote: | | ciaranm wrote: | | cokehabit wrote: | | paludis --isnt needed |
Ah, that's where you're so very wrong. Who else is going to come up with a solution to the "Portage doesn't do all the things developers need" problem? |
so paludis is for developers then |
Another is the package use masking and both kinds of use forcing |
Worth noting portage supports this already... |
Was wondering when you'd show up and begin with the misleading statements. |
|
| Back to top |
|
|
ferringb
Developer
Joined: 03 Apr 2003
Posts: 355
Location: USA
|
| Posted: Wed Sep 13, 2006 8:44 pm Post subject: |
|
|
| ciaranm wrote: | | ferringb wrote: | | ciaranm wrote: |
Another is the package use masking and both kinds of use forcing |
Worth noting portage supports this already... |
Was wondering when you'd show up and begin with the misleading statements. |
Not familiar with the "watch thread" option I take it.
Portage grew support for package.use.mask and use.force; that's why they're in the tree. Kind of hard to argue that is misleading, or you're just being snippy per the norm 
_________________
I don't want to be buried in a pet cemetery. ~Ramones |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Wed Sep 13, 2006 8:51 pm Post subject: |
|
|
| ferringb wrote: | | Portage grew support for package.use.mask and use.force; that's why they're in the tree. Kind of hard to argue that is misleading, or you're just being snippy per the norm |
So, the full story... Unstable Portage grew support for them several months after Paludis did, despite your protests, and anyone using them in the tree is subjected to your wrath. |
|
| Back to top |
|
|
ferringb
Developer
Joined: 03 Apr 2003
Posts: 355
Location: USA
|
| Posted: Wed Sep 13, 2006 8:58 pm Post subject: |
|
|
| ciaranm wrote: | | ferringb wrote: | | Portage grew support for package.use.mask and use.force; that's why they're in the tree. Kind of hard to argue that is misleading, or you're just being snippy per the norm |
So, the full story... Unstable Portage grew support for them several months after Paludis did, despite your protests, and anyone using them in the tree is subjected to your wrath. |
Bah, who is misleading now?
My protest was over the format; would rather see use dep syntax used for this, but TGL pointed out a flaw in it, and I shut up on that angle.
My wraith is reserved for people using features in profiles that require unstable portage without the dep; it's called QA, albeit in some cases not applicable. It's actually the same rules for _any_ change to the profiles; deploy stable support for 6 months, or level appropriate dep.
Like it or not, it's one aspect of QA most folks ignore- we wind up getting fugly messes like the .50->.51 cascaded profiles change over, or virtuals/* when folks violate it.
In regards to paludis having it prior, made no comment- so... in short, wasn't misleading.
Portage has support for it.
Also doesn't require you being a combative lil bugger either I might add
_________________
I don't want to be buried in a pet cemetery. ~Ramones |
|
| Back to top |
|
|
nesl247
Veteran
Joined: 15 Jun 2004
Posts: 1614
Location: Florida
|
| Posted: Sun Sep 17, 2006 2:12 am Post subject: |
|
|
ciaranm I have a few questions for you.
Will you support the folder structure for configuration (like portage-2.1)
How do I disable all that annoying test every time I run paludis -i?
Why do I have to manually create /var/paludis/repositories/*/profiles/categories? - it works just fine as long even if it's blank
Why do I have to specify x- in front of repository names when I manually sync them?
Why isn't there a generic /var/paludis/distfiles directory instead of putting it in ${location}/distfiles by default?
Will there ever be a paludis --ask or -a like in portage? - This is my major gripe
Why are the files called package_unmask and package_mask when use.conf and keywords.conf don't have package_ in the name?
How do I read the news as I don't see an option to, only to update it?
This about sums up the things I'm wondering and dislike. Other than that, great job because this is alot faster than portage. |
|
| Back to top |
|
|
ciaranm
Retired Dev
Joined: 19 Jul 2003
Posts: 1719
Location: In Hiding
|
| Posted: Sun Sep 17, 2006 2:23 am Post subject: |
|
|
| nesl247 wrote: | | Will you support the folder structure for configuration (like portage-2.1) |
No. Our configuration structure is nothing like Portage's.
| Quote: | | How do I disable all that annoying test every time I run paludis -i? |
Play with --log-level. Better yet, fix the errors.
| Quote: | | Why do I have to manually create /var/paludis/repositories/*/profiles/categories? - it works just fine as long even if it's blank |
Doesn't work just fine if it's blank. Paludis doesn't use overlays, it uses full multiple repository support. As such, each profile needs to provide its own full categories file.
| Quote: | | Why do I have to specify x- in front of repository names when I manually sync them? |
Because your repositories don't have a repo_name file like they should.
| Quote: | | Why isn't there a generic /var/paludis/distfiles directory instead of putting it in ${location}/distfiles by default? |
To avoid conflicts. If you want to take the risk, you're free to override it (and it'll be even easier to do so whenever someone gets around to making a repo defaults file).
| Quote: | | Will there ever be a paludis --ask or -a like in portage? - This is my major gripe |
The standard Paludis client will always be entirely non-interactive. Other clients, like gtkpaludis, can behave differently.
| Quote: | | Why are the files called package_unmask and package_mask when use.conf and keywords.conf don't have package_ in the name? |
To distinguish between package masks and use masks.
| Quote: | | How do I read the news as I don't see an option to, only to update it? |
eselect news |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Off the Wall
