Hi All,
I noticed Intel has announced [1] the first alpha of Moblin 2 which from memory and by the look of it is based on Fedora [2].
I'm wondering what the outcome of the discussions with them was about basing it on upstream Fedora and if anyone knows what the differences are between the source repos [2] and that of base Fedora are?
Cheers, Peter
[1] http://moblin.org/community/blogs/tshureih/2009/announcing-moblin-v2-core-al... [2] http://repo.moblin.org/moblin/releases/test/alpha1/
Peter Robinson wrote:
Hi All,
I noticed Intel has announced [1] the first alpha of Moblin 2 which from memory and by the look of it is based on Fedora [2].
Yes but Intel doesn't seem interested in pushing any changes back to Fedora.
I'm wondering what the outcome of the discussions with them was about basing it on upstream Fedora and if anyone knows what the differences are between the source repos [2] and that of base Fedora are?
No idea but have you tried asking in the Moblin mailing lists? They should know better.
Rahul
Hi All,
I noticed Intel has announced [1] the first alpha of Moblin 2 which from memory and by the look of it is based on Fedora [2].
Yes but Intel doesn't seem interested in pushing any changes back to Fedora.
I'm wondering what the outcome of the discussions with them was about basing it on upstream Fedora and if anyone knows what the differences are between the source repos [2] and that of base Fedora are?
No idea but have you tried asking in the Moblin mailing lists? They should know better.
The reason I asked was because there was a discussion about on fedora-devel it back when it was first announced that it would be based on Fedora. I seem to remember one of the board members was at the event where it was announced and was going to chat with them or something.
Peter
On Sat, 31 Jan 2009 18:57:42 +0100 Peter Robinson pbrobinson@gmail.com wrote:
Hi All,
I noticed Intel has announced [1] the first alpha of Moblin 2 which from memory and by the look of it is based on Fedora [2].
Various parts from Moblin 2 come from Fedora. Other parts come from OpenSUSE, and about half is packaged newly based on upstream releases. (although some descriptions are borrowed from various sources)
Yes but Intel doesn't seem interested in pushing any changes back to Fedora.
If there was a "based on" relationship we would. But it's not nearly so simple.
On Sun, 01 Feb 2009 02:32:25 +0530 Rahul Sundaram sundaram@fedoraproject.org wrote:
Arjan van de Ven wrote:
If there was a "based on" relationship we would. But it's not nearly so simple.
That would depend on what of Moblin came from Fedora and what changes Intel made. I wouldn't dismiss the possibility so easily.
I don't dismiss anything beforehand, and am quite open to discuss just about anything.
But frankly, if I look at where things are today in Moblin and where they are going; the packages we borrowed from Fedora tend to be those where we don't do much if any changes. If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
Arjan van de Ven wrote:
If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
How about a working version of the Poulsbo video driver?
See Adam Williamson's rant: http://www.happyassassin.net/2009/01/30/intel-gma-500-poulsbo-graphics-on-li...
Kevin Kofler
On Sun, 01 Feb 2009 00:47:44 +0100 Kevin Kofler kevin.kofler@chello.at wrote:
Arjan van de Ven wrote:
If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
How about a working version of the Poulsbo video driver?
we currently do not have one, as our release notes mention.
Arjan van de Ven wrote:
If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
Can we put improved boot time on the wish list? Whatever is reasonable and practical. Though Im not suggesting the expectation should be 5 second boot time for Fedora.
Brian
On Saturday 31 January 2009 05:07:45 pm Brian Maly wrote:
Arjan van de Ven wrote:
If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
Can we put improved boot time on the wish list? Whatever is reasonable and practical. Though Im not suggesting the expectation should be 5 second boot time for Fedora.
Brian
http://fedoraproject.org/wiki/Features/20SecondStartup
On Sat, 31 Jan 2009 20:07:45 -0500 Brian Maly bmaly@redhat.com wrote:
Arjan van de Ven wrote:
If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
Can we put improved boot time on the wish list? Whatever is reasonable and practical. Though Im not suggesting the expectation should be 5 second boot time for Fedora.
arguably that's clearly way outside of where Moblin borrowed some packages from Fedora ;-)
I have offered before, and that offer still stands, to help in this regard. And I have been working with various folks from Fedora (as well as other distros) on this.
It's not like there's any secrets; the LPC presentation I think lines out pretty well what process worked for us. It's just that the project will/motivation needs to be there to make this an important item, and then move relatively fast across a wide range of things to make them not-suck.
On Sat, Jan 31, 2009 at 2:34 PM, Arjan van de Ven arjan@infradead.org wrote:
But frankly, if I look at where things are today in Moblin and where they are going; the packages we borrowed from Fedora tend to be those where we don't do much if any changes. If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
Hmm, is there a good way to generate a list of changes between moblin and Fedora packages that would help us identify some specific things we could get back? Is this a matter of cataloging the Moblin specific patch files in the Moblin srpms?
The other question is, does the way Moblin "borrow" packages from Fedora fall under the newish Fedora Remix trademark guidance? And if so would Intel be willing to advertise (in some fashion) Moblin as a Fedora remix?
-jef
On Mon, 2 Feb 2009 07:25:37 -0900 Jeff Spaleta jspaleta@gmail.com wrote:
Hmm, is there a good way to generate a list of changes between moblin and Fedora packages that would help us identify some specific things we could get back? Is this a matter of cataloging the Moblin specific patch files in the Moblin srpms?
as I said before, the things we borrow we don't tend to patch
The other question is, does the way Moblin "borrow" packages from Fedora fall under the newish Fedora Remix trademark guidance?
I doubt it. I really don't consider Moblin "fedora + changes". It really isn't. It is much more a "New OS + borrow from Fedora and SuSE" I suspect (I haven't done the count) that less than half the packages fall under the "borrow from Fedora" umbrella. Probably more if you don't count applications.
And if so would Intel be willing to advertise (in some fashion) Moblin as a Fedora remix?
I don't think that is currently a relevant question...
On Mon, 2009-02-02 at 10:10 -0800, Arjan van de Ven wrote:
I doubt it. I really don't consider Moblin "fedora + changes". It really isn't. It is much more a "New OS + borrow from Fedora and SuSE" I suspect (I haven't done the count) that less than half the packages fall under the "borrow from Fedora" umbrella. Probably more if you don't count applications.
This does bring up an interesting question. What did you have to borrow from SuSE that the Fedora equiv (if it existed) wasn't suitable enough?
On Mon, 02 Feb 2009 10:13:03 -0800 Jesse Keating jkeating@redhat.com wrote:
On Mon, 2009-02-02 at 10:10 -0800, Arjan van de Ven wrote:
I doubt it. I really don't consider Moblin "fedora + changes". It really isn't. It is much more a "New OS + borrow from Fedora and SuSE" I suspect (I haven't done the count) that less than half the packages fall under the "borrow from Fedora" umbrella. Probably more if you don't count applications.
This does bring up an interesting question. What did you have to borrow from SuSE that the Fedora equiv (if it existed) wasn't suitable enough?
for example we took the entire toolchain from SuSE.
this is not a judgment about the Fedora toolchain. Just we decided on using the SuSE one. **This not an "instead of"**.
Your question only makes sense if you have the idea that we're doing a modified Fedora. We do not have that goal or idea. I realize there was some press earlier that implied that, but don't believe everything the press says.
In addition to the fedora or suse originated packages there are many packages that aren't from either of them, and I expect this to grow not shrink....
So I would like to really ask you and others to stop thinking of Moblin as "Fedora with changes" and measure everything against that. I realize it's easy to think that, and a lot of things just won't make sense in that mindset.
On Mon, Feb 2, 2009 at 10:35 AM, Arjan van de Ven arjan@infradead.org wrote:
So I would like to really ask you and others to stop thinking of Moblin as "Fedora with changes" and measure everything against that. I realize it's easy to think that, and a lot of things just won't make sense in that mindset.
But, I even think of Debian as Fedora with changes :->
-jef
On Mon, 2009-02-02 at 11:35 -0800, Arjan van de Ven wrote:
for example we took the entire toolchain from SuSE.
this is not a judgment about the Fedora toolchain. Just we decided on using the SuSE one. **This not an "instead of"**.
Your question only makes sense if you have the idea that we're doing a modified Fedora. We do not have that goal or idea. I realize there was some press earlier that implied that, but don't believe everything the press says.
In addition to the fedora or suse originated packages there are many packages that aren't from either of them, and I expect this to grow not shrink....
So I would like to really ask you and others to stop thinking of Moblin as "Fedora with changes" and measure everything against that. I realize it's easy to think that, and a lot of things just won't make sense in that mindset.
I think I'm just trying to wrap my head around "We're going to use some bits from Fedora, and some bits from SuSE, with no real reason why, and we just hope they'll work together. Should be fun!"
If I were setting out to make a new OS and I was going to borrow bits from folks, I'd try to single source all my bits possible to reduce any self inflicted integration issues. I wouldn't try to bolt together Chevy parts, Honda parts, and forge some parts on my own, that just doesn't seem like fun.
Look, I"m happy that we're able to provide some bits to you, that's one of the reasons why we are here. But there really must be a decision factor somewhere that would explain why some bits are ours, and some bits are others, and I'd like to know what that is, so that the next group that comes along doesn't have to look beyond what Fedora offers when they want to build their own OS.
On Mon, Feb 02, 2009 at 11:35:24AM -0800, Arjan van de Ven wrote:
So I would like to really ask you and others to stop thinking of Moblin as "Fedora with changes" and measure everything against that. I realize it's easy to think that, and a lot of things just won't make sense in that mindset.
I ended up analysing this while otherwise bored. There's nothing especially surprising in the moblin repositories. The large majority of the packages are Fedora derived, with a small number from suse (primarily the toolchain, as Arjan said) and a few custom ones.
Figuring out the proportion of the packages that were Fedora derived was actually surprisingly difficult. A large number of the specfiles have been processed through something called specbuilder. The behaviour of this seems to have varied between versions - some remove the original changelog, some don't. In some cases the specfiles are identical to the Fedora ones (to the extent of including comments) but have simply had the changelog entries stripped.
Using a few heuristics, I'd estimate that somewhere in the region of 80 packages (ie, <10%) of moblin is from-scratch packaging by Intel. The rest is either obtained from Suse (gcc, cache, gdbm, gmp, osc - binutils is from Fedora, oddly), is identical to the Fedora package or is a simple mechanical transformation of a Fedora package, in some cases updated to a newer upstream as compared to the point where the moblin packages were forked.
Moblin includes some 2000 patches, and again most of these are derived from Fedora or Suse. Moblin-specific patches are mostly either backports or code that has been submitted upstream but not yet released. There's a few counterexamples (code that should be upstream but doesn't seem to have been submitted), but that's in the minority and certainly no worse than any other distribution. The rest are changes to defaults or UI alterations to fit in with limited screen resolutions.
There's absolutely nothing wrong with any of this, but right now it's kind of hard to see how moblin is anyone other than Fedora with changes. I don't think that puts Intel under any sort of obligation to feed changes back to us and I agree that Koji isn't ideally suited to producing the kind of derivative that Intel want to, but it would be nice to acknowledge the amount of the project that's built on the work of Fedora contributors.
On Sun, May 3, 2009 at 8:08 AM, Matthew Garrett mjg@redhat.com wrote:
There's absolutely nothing wrong with any of this, but right now it's kind of hard to see how moblin is anyone other than Fedora with changes. I don't think that puts Intel under any sort of obligation to feed changes back to us and I agree that Koji isn't ideally suited to producing the kind of derivative that Intel want to, but it would be nice to acknowledge the amount of the project that's built on the work of Fedora contributors.
There is more than a touch of irony here, as in the past Fedora packagers have been on occasion have ruffled feathers of 3rd party by using 3rd party specfiles as a starting point for a package view for inclusion into Fedora. We certainly aren't in a position to demand any sort of official recognition. But at the same time we aren't restrained from point out that our process is influential.
I guess ideally I'd like to see it work like how rock bands influence each other. To be an influential rock band, that inherently means that other bands come up behind you and take your style of doing things along with pieces of other bands that have influenced them and make something new. The more influential a rock band you are, the more rock bands you influence. Doesn't mean you were packing football stadiums with fans..it means the other artists and the potential artists..the creators of value..were directly impacted by your work and incorporated it into their own. And of course those newer rock bands don't spend all that much time making it a point that they were influenced by other bands. Sure if they are directly asked in an interview they'll gladly talk about it..but in general they don't spend 15 minutes in their shows stepping through their evolution as artists. The fans that care can hear the influences in the new band's music. The question is, can moblin contributors hear/feel/smell/taste the influence from Fedora in the work moblin is doing?
-jef"So basically what I'm saying is.... maybe Fedora is open source's The Ramones... and moblin is headed towards being U2 or maybe Greenday."Spaleta
2009/5/3 Jeff Spaleta jspaleta@gmail.com:
On Sun, May 3, 2009 at 8:08 AM, Matthew Garrett mjg@redhat.com wrote:
There's absolutely nothing wrong with any of this, but right now it's kind of hard to see how moblin is anyone other than Fedora with changes. I don't think that puts Intel under any sort of obligation to feed changes back to us and I agree that Koji isn't ideally suited to producing the kind of derivative that Intel want to, but it would be nice to acknowledge the amount of the project that's built on the work of Fedora contributors.
There is more than a touch of irony here, as in the past Fedora packagers have been on occasion have ruffled feathers of 3rd party by using 3rd party specfiles as a starting point for a package view for inclusion into Fedora. We certainly aren't in a position to demand any sort of official recognition. But at the same time we aren't restrained from point out that our process is influential.
I guess ideally I'd like to see it work like how rock bands influence each other. To be an influential rock band, that inherently means that other bands come up behind you and take your style of doing things along with pieces of other bands that have influenced them and make something new. The more influential a rock band you are, the more rock bands you influence. Doesn't mean you were packing football stadiums with fans..it means the other artists and the potential artists..the creators of value..were directly impacted by your work and incorporated it into their own. And of course those newer rock bands don't spend all that much time making it a point that they were influenced by other bands. Sure if they are directly asked in an interview they'll gladly talk about it..but in general they don't spend 15 minutes in their shows stepping through their evolution as artists. The fans that care can hear the influences in the new band's music. The question is, can moblin contributors hear/feel/smell/taste the influence from Fedora in the work moblin is doing?
-jef"So basically what I'm saying is.... maybe Fedora is open source's The Ramones... and moblin is headed towards being U2 or maybe Greenday."Spaleta
The influences analogy is not one I would have chosen. U2 don't play "Sheena is a punk rocker" at their gigs (with a few different chords (patches)) and pretend it is theirs.
Its not quite Oracle Unbreakable though, which is nice.
I think really what has been identified are two things that Fedora would like to see.
1. Patches 2. Attribution
With reference to 1, Arjan has stated that some are so Moblin specific there is little point trying to upstream them. This happens in Fedora too. Some patches are so Fedora-specific that there is no point trying to get them applied. What I think people are saying is that Moblin could be doing more. Perhaps both Fedora and Moblin could consider splitting out patches into two categories:
fedora-foobarchanges.patch foobarchanges.patch
Those that start fedora- are stated as "There is never any intention to get this stuff upstream as it won't benefit anybody, its stuff that $PROJECT requires but no-one else". Then have some simple justification in the spec file.
With reference to 2, I believe it simply courteous of projects downstream to mention those on whose shoulders they stand in whatever manner they deem fit.
Regards
On Sun, 2009-05-03 at 17:08 +0100, Matthew Garrett wrote:
On Mon, Feb 02, 2009 at 11:35:24AM -0800, Arjan van de Ven wrote:
So I would like to really ask you and others to stop thinking of Moblin as "Fedora with changes" and measure everything against that. I realize it's easy to think that, and a lot of things just won't make sense in that mindset.
I ended up analysing this while otherwise bored. There's nothing especially surprising in the moblin repositories. The large majority of the packages are Fedora derived, with a small number from suse (primarily the toolchain, as Arjan said) and a few custom ones.
Figuring out the proportion of the packages that were Fedora derived was actually surprisingly difficult. A large number of the specfiles have been processed through something called specbuilder. The behaviour of this seems to have varied between versions - some remove the original changelog, some don't. In some cases the specfiles are identical to the Fedora ones (to the extent of including comments) but have simply had the changelog entries stripped.
There is a reasonably legitimate technical reason for this, which is that the changelogs end up in the rpmdb, which is wasted disk space for the moblin use case (user just wants firefox and doesn't care about package changelogs). It's about 24M on my machine, for instance.
Of course that's also something you could strip out at rpmbuild time...
There's absolutely nothing wrong with any of this, but right now it's kind of hard to see how moblin is anyone other than Fedora with changes. I don't think that puts Intel under any sort of obligation to feed changes back to us and I agree that Koji isn't ideally suited to producing the kind of derivative that Intel want to, but it would be nice to acknowledge the amount of the project that's built on the work of Fedora contributors.
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source: http://moblin.org/documentation/moblin-overview/faq ]
This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
- ajax
So I would like to really ask you and others to stop thinking of Moblin as "Fedora with changes" and measure everything against that. I realize it's easy to think that, and a lot of things just won't make sense in that mindset.
I ended up analysing this while otherwise bored. There's nothing especially surprising in the moblin repositories. The large majority of the packages are Fedora derived, with a small number from suse (primarily the toolchain, as Arjan said) and a few custom ones.
Figuring out the proportion of the packages that were Fedora derived was actually surprisingly difficult. A large number of the specfiles have been processed through something called specbuilder. The behaviour of this seems to have varied between versions - some remove the original changelog, some don't. In some cases the specfiles are identical to the Fedora ones (to the extent of including comments) but have simply had the changelog entries stripped.
There is a reasonably legitimate technical reason for this, which is that the changelogs end up in the rpmdb, which is wasted disk space for the moblin use case (user just wants firefox and doesn't care about package changelogs). It's about 24M on my machine, for instance.
Of course that's also something you could strip out at rpmbuild time...
There's absolutely nothing wrong with any of this, but right now it's kind of hard to see how moblin is anyone other than Fedora with changes. I don't think that puts Intel under any sort of obligation to feed changes back to us and I agree that Koji isn't ideally suited to producing the kind of derivative that Intel want to, but it would be nice to acknowledge the amount of the project that's built on the work of Fedora contributors.
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source: http://moblin.org/documentation/moblin-overview/faq ]
This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
And why when moblin 2 was announced did they make announcements that they were moving from Ubuntu to Fedora such as reported here http://www.theregister.co.uk/2008/07/23/moblin_reworked/
I would also be interested to know what was wrong with the Fedora toolchain that they decided to use the Suse one (or maybe what was right about the suse one).
Peter
On 05/04/2009 08:28 PM, Adam Jackson wrote:
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source: http://moblin.org/documentation/moblin-overview/faq ]This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
I don't see us accomplishing much by stating the obvious, which is that Moblin is indeed based on Fedora even if Intel does not want to acknowledge that for whatever reasons. Considering that they seem to have moved it over to Linux Foundation, it all might just be political considerations. Let's move beyond that.
Now, is there useful patches that we need to push upstream? Are there additional packages, we can import into Fedora? Let's look at that list. We know of sreadahead. Has the kernel portion been upstreamed? Arjan pointed out memuse in http://lwn.net/Articles/331423/. What's the rest?
Rahul
On Mon, 2009-05-04 at 21:01 +0530, Rahul Sundaram wrote:
On 05/04/2009 08:28 PM, Adam Jackson wrote:
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source: http://moblin.org/documentation/moblin-overview/faq ]This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
I don't see us accomplishing much by stating the obvious, which is that Moblin is indeed based on Fedora even if Intel does not want to acknowledge that for whatever reasons. Considering that they seem to have moved it over to Linux Foundation, it all might just be political considerations. Let's move beyond that.
Now, is there useful patches that we need to push upstream? Are there additional packages, we can import into Fedora? Let's look at that list. We know of sreadahead. Has the kernel portion been upstreamed? Arjan pointed out memuse in http://lwn.net/Articles/331423/. What's the rest?
Yeah, how about Poulsbo support? Is anyone at Intel actually working on upstreaming the unencumbered 2D parts of that, including the kernel bits and the X driver? Random crack in gregkh's tree doesn't count.
Dan
On Mon, 2009-05-04 at 14:33 -0400, Dan Williams wrote:
Yeah, how about Poulsbo support? Is anyone at Intel actually working on upstreaming the unencumbered 2D parts of that, including the kernel bits and the X driver? Random crack in gregkh's tree doesn't count.
Well, on that topic, I notice a huge new pile of crack showed up in the Ubuntu Mobile special-sauce repositories on April 30, which appears to be the psb driver for Ubuntu 8.10. At least...
http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/x/xserver-xorg-video... http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/x/xpsb-glx/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/libd/libdrm-poulsbo/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/p/psb-kernel-source/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/p/psb-firmware/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/p/psb-meta/
oh, so much crack. Why must there be so much crack?!
so, yeah, it seems like at least one bit of Intel is still working on special sauce for Ubuntu (presumably at Dell's behest), not upstreaming.
On Mon, 2009-05-04 at 12:26 -0700, Adam Williamson wrote:
On Mon, 2009-05-04 at 14:33 -0400, Dan Williams wrote:
Yeah, how about Poulsbo support? Is anyone at Intel actually working on upstreaming the unencumbered 2D parts of that, including the kernel bits and the X driver? Random crack in gregkh's tree doesn't count.
Well, on that topic, I notice a huge new pile of crack showed up in the Ubuntu Mobile special-sauce repositories on April 30, which appears to be the psb driver for Ubuntu 8.10. At least...
http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/x/xserver-xorg-video... http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/x/xpsb-glx/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/libd/libdrm-poulsbo/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/p/psb-kernel-source/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/p/psb-firmware/ http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/p/psb-meta/
oh, so much crack. Why must there be so much crack?!
so, yeah, it seems like at least one bit of Intel is still working on special sauce for Ubuntu (presumably at Dell's behest), not upstreaming.
Yeah, and there's the problem. What makes the Poulsbo team so special that they are exempt from the upstreaming policy that every other part of Intel seems to follow so well these days? ...
Dan
On Mon, 04 May 2009 14:33:28 -0400 Dan Williams dcbw@redhat.com wrote:
On Mon, 2009-05-04 at 21:01 +0530, Rahul Sundaram wrote:
On 05/04/2009 08:28 PM, Adam Jackson wrote:
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source:http://moblin.org/documentation/moblin-overview/faq ]
This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
I don't see us accomplishing much by stating the obvious, which is that Moblin is indeed based on Fedora even if Intel does not want to acknowledge that for whatever reasons. Considering that they seem to have moved it over to Linux Foundation, it all might just be political considerations. Let's move beyond that.
Now, is there useful patches that we need to push upstream? Are there additional packages, we can import into Fedora? Let's look at that list. We know of sreadahead. Has the kernel portion been upstreamed? Arjan pointed out memuse in http://lwn.net/Articles/331423/. What's the rest?
Yeah, how about Poulsbo support? Is anyone at Intel actually working on upstreaming the unencumbered 2D parts of that, including the kernel bits and the X driver? Random crack in gregkh's tree doesn't count.
it's not very useful if the various upstream maintainers say that they won't accept it no matter what... at that point... yes people stop working on it.
On Mon, 2009-05-04 at 12:52 -0700, Arjan van de Ven wrote:
it's not very useful if the various upstream maintainers say that they won't accept it no matter what... at that point... yes people stop working on it.
I haven't seen anywhere where that's been said. On the kernel/drm side of things, the LKML discussion resulted in some questions about the large amount of proprietary code the changes seemed to depend on, but no definitive Yes or No was reached: Greg decided himself to change approach and try to come up with a basic 2D-only driver which would not depend on any proprietary code, then propose whatever kernel elements would be needed for that instead. AFAIK he is still working on that.
To my knowledge there hasn't been any kind of proposal of any Poulsbo-related code to X.org.
(and since Greg is working off his own bat on this - obviously not representing Intel, Canonical or Dell, or even Novell - there's been a sum total of zero attempts to upstream any Poulsbo-related code from anyone who's being paid by anyone to work on it...)
On Mon, 2009-05-04 at 12:52 -0700, Arjan van de Ven wrote:
On Mon, 04 May 2009 14:33:28 -0400 Dan Williams dcbw@redhat.com wrote:
On Mon, 2009-05-04 at 21:01 +0530, Rahul Sundaram wrote:
On 05/04/2009 08:28 PM, Adam Jackson wrote:
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source:http://moblin.org/documentation/moblin-overview/faq ]
This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
I don't see us accomplishing much by stating the obvious, which is that Moblin is indeed based on Fedora even if Intel does not want to acknowledge that for whatever reasons. Considering that they seem to have moved it over to Linux Foundation, it all might just be political considerations. Let's move beyond that.
Now, is there useful patches that we need to push upstream? Are there additional packages, we can import into Fedora? Let's look at that list. We know of sreadahead. Has the kernel portion been upstreamed? Arjan pointed out memuse in http://lwn.net/Articles/331423/. What's the rest?
Yeah, how about Poulsbo support? Is anyone at Intel actually working on upstreaming the unencumbered 2D parts of that, including the kernel bits and the X driver? Random crack in gregkh's tree doesn't count.
it's not very useful if the various upstream maintainers say that they won't accept it no matter what... at that point... yes people stop working on it.
http://marc.info/?l=linux-kernel&m=113378006232564&w=2
You've come a long way :)
Dave.
On 05/05/2009 01:22 AM, Arjan van de Ven wrote:
it's not very useful if the various upstream maintainers say that they won't accept it no matter what... at that point... yes people stop working on it.
Intel is the largest contributor to Xorg and only one ahead of Red Hat currently and it is no stranger to the Linux Kernel either. If it cannot get rid of the binary driver (remember your person stand on that?), replace it with a free and open source one and get it merged upstream, that's a damn shame.
When anyone asked me about what to look for, I always said, if your system comes with a Intel chipset, you are fine. Now, I cannot say that anymore and that is a disappointment.
Rahul
On Mon, 2009-05-04 at 12:52 -0700, Arjan van de Ven wrote:
On Mon, 04 May 2009 14:33:28 -0400 Dan Williams dcbw@redhat.com wrote:
On Mon, 2009-05-04 at 21:01 +0530, Rahul Sundaram wrote:
On 05/04/2009 08:28 PM, Adam Jackson wrote:
Just to underline this point, let's look at what the Moblin FAQ has to say on the subject:
Q Is Moblin v2.0 based on another distribution? Moblin v2.0 borrows components from various distributions, but is not based on another distribution.
[ source:http://moblin.org/documentation/moblin-overview/faq ]
This seems... disingenuous? I guess it all depends on what the definition of the word "based" is. It's also the sort of statement that begs immediate deconstruction. If moblin _isn't_ based on another distribution, why does it feel the need to say so. On the other hand, if it is, why does it say it isn't.
I don't see us accomplishing much by stating the obvious, which is that Moblin is indeed based on Fedora even if Intel does not want to acknowledge that for whatever reasons. Considering that they seem to have moved it over to Linux Foundation, it all might just be political considerations. Let's move beyond that.
Now, is there useful patches that we need to push upstream? Are there additional packages, we can import into Fedora? Let's look at that list. We know of sreadahead. Has the kernel portion been upstreamed? Arjan pointed out memuse in http://lwn.net/Articles/331423/. What's the rest?
Yeah, how about Poulsbo support? Is anyone at Intel actually working on upstreaming the unencumbered 2D parts of that, including the kernel bits and the X driver? Random crack in gregkh's tree doesn't count.
it's not very useful if the various upstream maintainers say that they won't accept it no matter what... at that point... yes people stop working on it.
Nobody has ever said that about the *2D* parts, and I specifically mentioned "unencumbered 2D parts" above. Intel needs to get its head out of its ass WRT Poulsbo and gets out some 2D bits, and stops working in private trees and stop doing code dumps only to Ubuntu.
Seriously, is that team even *part* of Intel? Because the Poulsbo Linux team is completely different than any of the other Intel Linux teams I've had the actual pleasure of interacting with.
Dan
On Thu, 2009-05-07 at 10:06 -0400, Dan Williams wrote:
Seriously, is that team even *part* of Intel? Because the Poulsbo Linux team is completely different than any of the other Intel Linux teams I've had the actual pleasure of interacting with.
At least originally, no, they weren't. Intel contracted the Poulsbo driver development out to another company (which has since itself been bought out by VMware). Now, however, it isn't clear if it's those guys or guys who really are with Intel or what who are doing whatever development is actually happening.
I suspect what may be happening is the original group is continuing to work on the crack-addled drivers that show up in the Ubuntu repos from time to time, while a different group works on creating a non-crack-addled driver that'll be sent through the proper processes. But that's just my guess, as no-one at Intel seems to want to say anything.
On 05/07/2009 08:58 PM, Adam Williamson wrote:
I suspect what may be happening is the original group is continuing to work on the crack-addled drivers that show up in the Ubuntu repos from time to time, while a different group works on creating a non-crack-addled driver that'll be sent through the proper processes. But that's just my guess, as no-one at Intel seems to want to say anything.
Whatever process that is, it is totally screwed up. The sooner they get off it and start following the successful upstream strategy they have so long, the better it is, for everyone.
Rahul
Seriously, is that team even *part* of Intel? Because the Poulsbo Linux team is completely different than any of the other Intel Linux teams I've had the actual pleasure of interacting with.
At least originally, no, they weren't. Intel contracted the Poulsbo driver development out to another company (which has since itself been bought out by VMware). Now, however, it isn't clear if it's those guys or guys who really are with Intel or what who are doing whatever development is actually happening.
I suspect what may be happening is the original group is continuing to work on the crack-addled drivers that show up in the Ubuntu repos from time to time, while a different group works on creating a non-crack-addled driver that'll be sent through the proper processes. But that's just my guess, as no-one at Intel seems to want to say anything.
Isn't the Poulsbo GPU based on a powerVR core that Intel licensed? That might explain some of the crack.
Peter
On Fri, 2009-05-08 at 09:21 +0100, Peter Robinson wrote:
I suspect what may be happening is the original group is continuing to work on the crack-addled drivers that show up in the Ubuntu repos from time to time, while a different group works on creating a non-crack-addled driver that'll be sent through the proper processes. But that's just my guess, as no-one at Intel seems to want to say anything.
Isn't the Poulsbo GPU based on a powerVR core that Intel licensed? That might explain some of the crack.
Yes, it more or less does; that's why they sub-contracted the development, the hardware is substantially different from previous Intel chips, and the existing Linux development team within Intel didn't have much familiarity with it, and was too busy to drop everything and learn about this new hardware. So they contracted development out to a group who are very familiar with the PowerVR hardware, but apparently think a driver which relies on a libdrm branch and three separate chunks of proprietary code (not to mention a badly-written kernel module just to make 2D work) is a sensible design.
I've been told, though, that the GMA500 is actually a Frankenstein design, with the PowerVR 3D (and hardware video playback acceleration) core glued on to a standard Intel 2D core, so if that's accurate, it should be reasonably easy to come with at least a basic native accelerated 2D driver which doesn't depend on all the horrible proprietary crack, which would be fine for a lot of people. The 3D and video playback acceleration features are nice, but not essential. It'd be nice if Intel would just kick out a driver like that, at least.
On Fri, 2009-05-08 at 11:03 -0700, Adam Williamson wrote:
Yes, it more or less does; that's why they sub-contracted the development, the hardware is substantially different from previous Intel chips, and the existing Linux development team within Intel didn't have much familiarity with it, and was too busy to drop everything and learn about this new hardware. So they contracted development out to a group who are very familiar with the PowerVR hardware, but apparently think a driver which relies on a libdrm branch and three separate chunks of proprietary code (not to mention a badly-written kernel module just to make 2D work) is a sensible design.
More accurately, that's what Tungsten _had_ to implement, because Imagination (the PowerVR people) were unwilling or unable to allow open code.
I've been told, though, that the GMA500 is actually a Frankenstein design, with the PowerVR 3D (and hardware video playback acceleration) core glued on to a standard Intel 2D core, so if that's accurate, it should be reasonably easy to come with at least a basic native accelerated 2D driver which doesn't depend on all the horrible proprietary crack, which would be fine for a lot of people. The 3D and video playback acceleration features are nice, but not essential. It'd be nice if Intel would just kick out a driver like that, at least.
IIRC it's an Intel-ish output block, but PVR acceleration blocks (both 2D and 3D). Not quite the same thing.
- ajax
On Fri, 2009-05-08 at 15:34 -0400, Adam Jackson wrote:
More accurately, that's what Tungsten _had_ to implement, because Imagination (the PowerVR people) were unwilling or unable to allow open code.
...
IIRC it's an Intel-ish output block, but PVR acceleration blocks (both 2D and 3D). Not quite the same thing.
Thanks for the clarifications.
On Monday 04 May 2009 17:58:35 Adam Jackson wrote:
On Sun, 2009-05-03 at 17:08 +0100, Matthew Garrett wrote:
been processed through something called specbuilder. The behaviour of this seems to have varied between versions - some remove the original changelog, some don't. In some cases the specfiles are identical to the Fedora ones (to the extent of including comments) but have simply had the changelog entries stripped.
There is a reasonably legitimate technical reason for this, which is that the changelogs end up in the rpmdb, which is wasted disk space for the moblin use case (user just wants firefox and doesn't care about package changelogs). It's about 24M on my machine, for instance.
Of course that's also something you could strip out at rpmbuild time...
Which sounds pretty good idea, so I went and asked:
#rpm.org <Tuju> hi, any change ever seeing that feature metioned in moblin thread in fedora-devel mailing list? it was about dropping changelog from binaries in rpmbuild. <PanuM> Tuju: that's actually already implemented upstream, just not in any released version yet <Tuju> well that's great news. <PanuM> here's the reference: http://rpm.org/ticket/47
DESCRIPTION To save space in the packages and the rpmdb a macro could limit the changelog in binary packages. That way ancient log entries could still be looked up in the spec file. Limit could be done by #entries or by time or both.
04/16/09 12:20:34 changed by jnovy status changed from new to closed. resolution set to fixed. The new changelog trimming feature is now implemented upstream. The current implementation allows changelog trimming based on changelog entry date. It is specified via _changelog_trimtime macro: # The Unix time of the latest kept changelog entry in binary packages. # Any older entry is not packaged in binary packages. %_changelog_trimtime 0
Tuju
On Sun, 3 May 2009 17:08:27 +0100 Matthew Garrett mjg@redhat.com wrote:
[ I wasn't on the CC, just saw this by accident]
A large number of the specfiles have been processed through something called specbuilder.
If there is anything in Moblin that I'd like Fedora to consider adopting, it is specbuilder. Specbuilder is basically the (start of) an autopackaging tool, where we try to automatically generate packaging for a class of upstream projects. The class we're focusing on is those upstream packages that use autoconf/automake/etc and use pkgconfig.
For these packages we have a very small .ini file with some basic metadata (which we're trying to autogenerate at much as possible as well), which specbuilder then turns into a .spec file for building.
These ini files are relatively small and simple, and can mostly be autogenerated. The result is a very consistent set of rpms, and we're working on being able to generate other packaging formats from this same minimal metadata as well....
For example the ini file for the Intel IIG graphics driver package looks like this:
[header] Summary = Xorg X11 Intel video driver Name = xorg-x11-drv-intel LocaleName = xorg-x11-drv-intel Version = 2.7.0 Release = 1 Group = User Interface/X Hardware Support License = MIT URL = http://www.x.org/ Files = xorg-x11-drv-intel.files Requires = xorg-x11-server-Xorg hwdata SubPackages = devel tools
[configuration] Sources = http://xorg.freedesktop.org/archive/individual/driver/xf86-video-intel-%%7Bv... PkgConfig = xorg-server libdrm xf86driproto xvmc gl x11 xproto fontsproto randrproto renderproto xextproto glproto xineramaproto xi dri2proto xext configure = reconfigure ConfigOptions = --enable-dri Patches = regdumper.patch copy-fb.patch biosdumper.patch
[devel] Summary = Xorg X11 Intel video driver XvMC development package Group = Development/System Files = xorg-x11-drv-intel-devel.files
[tools] Summary = Xorg X11 Intel video driver diagnostics tools Group = Development/System Files = xorg-x11-drv-intel-tools.files
On 05/04/2009 08:57 PM, Arjan van de Ven wrote:
On Sun, 3 May 2009 17:08:27 +0100 Matthew Garrett mjg@redhat.com wrote:
[ I wasn't on the CC, just saw this by accident]
A large number of the specfiles have been processed through something called specbuilder.
If there is anything in Moblin that I'd like Fedora to consider adopting, it is specbuilder. Specbuilder is basically the (start of) an autopackaging tool, where we try to automatically generate packaging for a class of upstream projects. The class we're focusing on is those upstream packages that use autoconf/automake/etc and use pkgconfig.
Is there a upstream project page or should we look directly at
http://repo.moblin.org/moblin/development/core/source/spec-builder-0.13-1.25... ?
Rahul
If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
Moblin Image Creater 2 [1] looks interesting to me. How does this compare to Fedora's own LiveCD tools given that it is based on them?
Cheerio, Debarshi
Arjan van de Ven wrote:
On Sun, 01 Feb 2009 02:32:25 +0530 Rahul Sundaram wrote:
Arjan van de Ven wrote:
If there was a "based on" relationship we would. But it's not nearly so simple.
That would depend on what of Moblin came from Fedora and what changes Intel made. I wouldn't dismiss the possibility so easily.
I don't dismiss anything beforehand, and am quite open to discuss just about anything.
But frankly, if I look at where things are today in Moblin and where they are going; the packages we borrowed from Fedora tend to be those where we don't do much if any changes. If there's something specific that the Fedora project wants to get from Moblin we should discuss that, but lets be specific there rather than having a discussion on the blanket hand-wavey level...
It has to remain hand-wavey until we get to know about what changes have been made. If there is no changes, then there is nothing further to discuss.
Rahul
Arjan van de Ven wrote:
On Sat, 31 Jan 2009 18:57:42 +0100 Peter Robinson pbrobinson@gmail.com wrote:
Hi All,
I noticed Intel has announced [1] the first alpha of Moblin 2 which from memory and by the look of it is based on Fedora [2].
Various parts from Moblin 2 come from Fedora. Other parts come from OpenSUSE, and about half is packaged newly based on upstream releases. (although some descriptions are borrowed from various sources)
Yes but Intel doesn't seem interested in pushing any changes back to Fedora.
If there was a "based on" relationship we would. But it's not nearly so simple.
For anyone interested in the gritty details, there is an article that explains the specifics of achieving the fast boot time. I presume the demo in question to be exactly whats discussed in the article.
LPC: Booting Linux in five seconds: http://lwn.net/Articles/299483/
Brian
devel@lists.stg.fedoraproject.org
-
Adam Jackson -
Adam Williamson -
Arjan van de Ven -
Brian Maly -
Christopher Brown -
Conrad Meyer -
Dan Williams -
David Airlie -
Debarshi Ray -
Jef Spaleta -
Jesse Keating -
Juha Tuomala -
Kevin Kofler -
Matthew Garrett -
Peter Robinson -
Rahul Sundaram