[RFC] Moving away from 'common'
Locked
[RFC] Moving away from 'common'
 
|
Hi,
I'm tired of "Automatic update of common submodule". Of the last 10 commits in gst-openmax only 2 have been real commits. This is high volume noise. Imagine a truly modular system like Xorg, where modules in-tree (xf86-video-ati) don't have any extra rights than modules out-of-tree (xf86-video-avivo). There are many advantages but one of the main ones is that new development is easy, and moving new modules to in-tree is much easier. Is GStreamer that modular? Well, I can already see a difference; in-tree modules will get automatic common submodule updates, while out-of-tree wouldn't. I've suggested previously to remove this automatic update but nobody found my arguments compelling. So I'm taking the next step which is to question if gst 'common' makes any sense at all. For starters the name 'common' suggests that all GStreamer submodules *must* have all that stuff, but that is not true. In gst-openmax for example I doubt I'm using even 10% of 'common', and how many modules really make use of gst-x11.m4 anyway? So, just like gnome-common, and xorg-util, I'm suggesting we create a gst-util package that is installed just like any other package so we get rid of git submodules. This move would probably be a bit big so it might make sense to do it step by step. Here's a little something to begin with: http://github.com/felipec/gst-util Thoughts? -- Felipe Contreras ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
On Wed, Mar 11, 2009 at 10:13:48PM +0200, Felipe Contreras wrote:
> I'm tired of "Automatic update of common submodule". Of the last 10 > commits in gst-openmax only 2 have been real commits. This is high > volume noise. Feel free to remove gst-openmax from the common-update script. dave... ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
On Wed, Mar 11, 2009 at 11:13 PM, David Schleef <[hidden email]> wrote:
> On Wed, Mar 11, 2009 at 10:13:48PM +0200, Felipe Contreras wrote: >> I'm tired of "Automatic update of common submodule". Of the last 10 >> commits in gst-openmax only 2 have been real commits. This is high >> volume noise. > > Feel free to remove gst-openmax from the common-update script. Done. No comments on the proposal? -- Felipe Contreras ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
In reply to this post by David Schleef-2
On Wed, 2009-03-11 at 14:13 -0700, David Schleef wrote:
> On Wed, Mar 11, 2009 at 10:13:48PM +0200, Felipe Contreras wrote: > > I'm tired of "Automatic update of common submodule". Of the last 10 > > commits in gst-openmax only 2 have been real commits. This is high > > volume noise. > > Feel free to remove gst-openmax from the common-update script. Is that a nice way of saying "if you don't like the game, feel free to take your ball and leave"? -- "XML is like violence: if it doesn't solve your problem, you aren't using enough of it." - Chris Maden ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
In reply to this post by Felipe Contreras
On Thu, 2009-03-12 at 00:00 +0200, Felipe Contreras wrote:
> No comments on the proposal? Besides the resounding 'no' you got on IRC yesterday you mean? Anyway, -1 from me for the record. The current system works well enough for me, although I can certainly think of improvements on the git side that would make things a bit nicer (some of which are being worked on already afaik). Cheers -Tim ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
In reply to this post by Stuart Jansen-2
On Wed, Mar 11, 2009 at 04:25:11PM -0600, Stuart Jansen wrote:
> On Wed, 2009-03-11 at 14:13 -0700, David Schleef wrote: > > On Wed, Mar 11, 2009 at 10:13:48PM +0200, Felipe Contreras wrote: > > > I'm tired of "Automatic update of common submodule". Of the last 10 > > > commits in gst-openmax only 2 have been real commits. This is high > > > volume noise. > > > > Feel free to remove gst-openmax from the common-update script. > > Is that a nice way of saying "if you don't like the game, feel free to > take your ball and leave"? Heh, no. The purpose of updating common everywhere is to make sure that any regressions are noticed quickly. If gst-openmax isn't (as) actively developed like, say, gst-plugins-good, then regressions probably *won't* get noticed quickly, and it would be better for a developer to update common manually. dave... ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
On Thu, Mar 12, 2009 at 1:08 AM, David Schleef <[hidden email]> wrote:
> On Wed, Mar 11, 2009 at 04:25:11PM -0600, Stuart Jansen wrote: >> On Wed, 2009-03-11 at 14:13 -0700, David Schleef wrote: >> > On Wed, Mar 11, 2009 at 10:13:48PM +0200, Felipe Contreras wrote: >> > > I'm tired of "Automatic update of common submodule". Of the last 10 >> > > commits in gst-openmax only 2 have been real commits. This is high >> > > volume noise. >> > >> > Feel free to remove gst-openmax from the common-update script. >> >> Is that a nice way of saying "if you don't like the game, feel free to >> take your ball and leave"? > > Heh, no. The purpose of updating common everywhere is to make sure > that any regressions are noticed quickly. If gst-openmax isn't (as) > actively developed like, say, gst-plugins-good, then regressions > probably *won't* get noticed quickly, and it would be better for a > developer to update common manually. I doubt the GStreamer project in general is interested in finding regressions quickly in 'common', regressions anywhere else are more important and are taking much more time right now. I find very low value in that. Also, I just did a git bisect in gst-openmax and again, automatic common updates just introduced noise. Annoying-productivity-decreasing noise. If somebody is interested in finding regressions in 'common' he can manually update common in all the repositories he is interested in, and test. Whenever there's a new feature in 'common' that person is interested in, he can commit the 'common' update in his favourite repos and make use of the feature. Otherwise I don't see any value in updating common. For example, somebody adds the 'shave' stuff into 'common'... should all the modules be updated because of that? No, only the ones that would actually make use of it, the automatic update is a complete waste of resources. Besides, if automatic submodule updates are so good how could Xorg survive without them? Are they finding regressions in xorg-util too late? So coming back to my proposal, your response suggests that detaching common would be bad because we will loose automatic updates. Is that the case? -- Felipe Contreras ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
In reply to this post by Felipe Contreras
Felipe Contreras wrote:
> Hi, > > I'm tired of "Automatic update of common submodule". Of the last 10 > commits in gst-openmax only 2 have been real commits. This is high > volume noise. [...] > So, just like gnome-common, and xorg-util, I'm suggesting we create a > gst-util package that is installed just like any other package so we > get rid of git submodules. > > This move would probably be a bit big so it might make sense to do it > step by step. Here's a little something to begin with: > http://github.com/felipec/gst-util > > Thoughts? > Just thinking out loud here, what about a hybrid approach (since the core developers seem to be so against letting go of git submodule): Maybe it's possible to add some stuff to the common module to make it installable, while keeping it as a submodule as well. Then nothing would change, gst core repos would keep their setup but in addition you could install gst-common itself to use it for out-of-tree modules as well. autogen.sh now checks out the submodule, if this could be changed to: if common/ exists -> git submodule update (for developing common itself) else if $(gst-common --version) >= $REQUIRED_COMMON -> use system common else git submodule init && git submodule update Just a thought. Would be nice to have because it would make everybody happy AFAICS. -- Regards, René Stadler ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
In reply to this post by Tim-Philipp Müller-2
On Wed, 2009-03-11 at 23:01 +0000, Tim-Philipp Müller wrote:
> On Thu, 2009-03-12 at 00:00 +0200, Felipe Contreras wrote: > > > No comments on the proposal? > > Besides the resounding 'no' you got on IRC yesterday you mean? > > Anyway, -1 from me for the record. The current system works well enough > for me, although I can certainly think of improvements on the git side > that would make things a bit nicer (some of which are being worked on > already afaik). > And from me. The fact that git can't accomodate the model we had with CVS is a failing of git. And one that they are apparently working on. I really dislike the idea of a gst-util module for build-time dependencies. J. -- Jan Schmidt <[hidden email]> ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
On Thu, Mar 12, 2009 at 1:09 PM, Jan Schmidt <[hidden email]> wrote:
> On Wed, 2009-03-11 at 23:01 +0000, Tim-Philipp Müller wrote: >> On Thu, 2009-03-12 at 00:00 +0200, Felipe Contreras wrote: >> >> > No comments on the proposal? >> >> Besides the resounding 'no' you got on IRC yesterday you mean? >> >> Anyway, -1 from me for the record. The current system works well enough >> for me, although I can certainly think of improvements on the git side >> that would make things a bit nicer (some of which are being worked on >> already afaik). >> > > And from me. The fact that git can't accomodate the model we had with > CVS is a failing of git. And one that they are apparently working on. Or a failure of the model. I keep hearing "they are working on it", I have been following git development since a long time now (even participated) and I only see sporadic patches for git submodule, but in my view it's still a second class citizen... nobody cares about it (or rather... very few people). > I really dislike the idea of a gst-util module for build-time > dependencies. So essentially you dislike having /usr/share/aclocal/*.m4 in your system. Following that mentality then /usr/share/aclocal/glib.m4 is bad, you should add a git submodule for glib-common. How about a plugin developer that doesn't like git and uses mercurial, or bzr? How are they supposed to have all the 'common' fanciness? -- Felipe Contreras ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: [RFC] Moving away from 'common'
|
|
In reply to this post by Tim-Philipp Müller-2
On Thu, Mar 12, 2009 at 1:01 AM, Tim-Philipp Müller <[hidden email]> wrote:
> On Thu, 2009-03-12 at 00:00 +0200, Felipe Contreras wrote: > >> No comments on the proposal? > > Besides the resounding 'no' you got on IRC yesterday you mean? I didn't propose this on IRC, and even if I did; IRC != GStreamer community, or rather... the guys on IRC on a particular time-zone. > Anyway, -1 from me for the record. The current system works well enough > for me, although I can certainly think of improvements on the git side > that would make things a bit nicer (some of which are being worked on > already afaik). What makes you think it's being worked on? It definitely can be improved if somebody who cares makes these improvements. Lets say for example; GStreamer developers. Who else needs git submodules? -- Felipe Contreras ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
«
Return to GStreamer-devel
|
1 view|%1 views
| Free forum by Nabble | Edit this page |