Hi all. It has been brought to my attention that my description of my future vision of rawhide as explained here is much clearer than previous attempts (including the current "no frozen rawhide" wiki page). So I felt it prudent to forward it along to the devel list for more eyes to look upon it and comment if desired.
Thanks!
-------- Forwarded Message -------- From: Jesse Keating jkeating@redhat.com Reply-to: fedora-advisory-board@redhat.com To: fedora-advisory-board@redhat.com Subject: Re: "What is the Fedora Project?" Date: Wed, 21 Oct 2009 18:58:08 -0700
On Wed, 2009-10-21 at 20:35 -0400, Paul W. Frields wrote:
I've heard a bit of preliminary rumbling about DSCM-like Rawhides -- a way for developers to have trees that move at their pace, and are possibly quite broken from time to time in ways that differ from each other. If we were able to develop such a scenario, why not also provide the flipside of this idea -- make the One True Rawhide the place where we take in changes that don't break the world, while they're cobbled on in the other trees? Whether this is an extension of the "KoPeR" idea or something entirely difficult, it merits serious consideration.
I very much like the aspect of the more stable rawhide here.
Jesse Keating brought up some concerns about integration, but aren't those concerns something that people would be interested in solving? (I'm assuming those people are the wide variety of engineers working in the Fedora community who are smarter than I.)
So my plans are really funny. I plan to make rawhide more unstable more of the time, and I plan to make "rawhide" more stable more of the time. Crazy eh? How can I do this? By splitting "rawhide" in two.
Rawhide as we know it, /pub/fedora/linux/releases/development/ will remain "rawhide". We may even change the path to say rawhide, just to catch things up and well I like keeping mirrors on their toes. Rawhide will be a repository of developmental and experimental packages. Things being worked on for the future. It will /not/ be an installable tree, rather it will just be a repository of packages, to be added on to an already stable "base", eg you'd install F12, and enable rawhide to test rawhide. This will significantly lower the complaints that "rawhide isn't installable".
The second face of rawhide, will be the "pending release", that is when it comes time to feature freeze a release, we'll split it away from rawhide. We'll publish to /pub/fedora/linux/releases/test/13-pending/ or some such. THIS tree will be installable. It will be composed each night, and we'll use bodhi to manage updates to this tree. That means this tree will have it's own "updates testing" where potential freeze breaks can be tested and commented on by all, but won't risk the base tree. If testing pans out, it'll get tagged for the release, if not it'll get thrown away. This tree will spawn 13-Alpha, 13-Beta, the snapshots in between, and eventually pub/fedora/linux/releases/13.
Remember that first rawhide? Yeah, it kept going, unfrozen, leaping toward Fedora 14. You could still install 13-Alpha, or 13-pending, and enable/update to rawhide to start testing Fedora 14 stuff.
What does this accomplish? It provides a very easy release valve. Instead of closing the valve and building up pressure while we freeze, and tempting people to push things into our pending release that really don't belong, we'll provide them a normal, never ending release of pressure, called rawhide. You can always find the latest stuff in rawhide, there is nothing newer (unless we make KoPeRs happen). We don't have to worry about "rawhide" being installable. We don't have to worry about people dumping highly experimental or developmental stuff in our pending release. We don't have to worry about the giant pile of builds for the next release building up while we polish the pending release. We don't have to worry about the giant pile of 0-day updates building up while we polish the pending release, as we'll be pushing these updates as we go.
This is my vision on how to accomplish both a always active development stream, and a more stable pending release stream, keeping everybody happy. Want to help? I'll be at FUDCon Toronto discussing roadblocks to this vision and discussing why this vision sucks if anybody thinks that it does. Or just find me on IRC/email if you want to chat about it.
_______________________________________________ fedora-advisory-board mailing list fedora-advisory-board@redhat.com http://www.redhat.com/mailman/listinfo/fedora-advisory-board
On Thu, 2009-10-22 at 09:50 -0700, Jesse Keating wrote:
This is my vision on how to accomplish both a always active development stream, and a more stable pending release stream, keeping everybody happy. Want to help? I'll be at FUDCon Toronto discussing roadblocks to this vision and discussing why this vision sucks if anybody thinks that it does. Or just find me on IRC/email if you want to chat about it.
Not a bad idea. In all honesty, I find rawhide is rarely fully installable out of the "box" anyway...this isn't a grumble! I think it's better to embrace reality and have multiple trees, though I am interested in whether that pending release couldn't also one day have a counterpart that had more longevity - a "testing" release (not the same as the "testing" that we currently have through updates) :)
Jon.
On Thu, 2009-10-22 at 09:50 -0700, Jesse Keating wrote:
Hi all. It has been brought to my attention that my description of my future vision of rawhide as explained here is much clearer than previous attempts (including the current "no frozen rawhide" wiki page). So I felt it prudent to forward it along to the devel list for more eyes to look upon it and comment if desired.
Thanks!
-------- Forwarded Message -------- From: Jesse Keating jkeating@redhat.com Reply-to: fedora-advisory-board@redhat.com To: fedora-advisory-board@redhat.com Subject: Re: "What is the Fedora Project?" Date: Wed, 21 Oct 2009 18:58:08 -0700
On Wed, 2009-10-21 at 20:35 -0400, Paul W. Frields wrote:
I've heard a bit of preliminary rumbling about DSCM-like Rawhides -- a way for developers to have trees that move at their pace, and are possibly quite broken from time to time in ways that differ from each other. If we were able to develop such a scenario, why not also provide the flipside of this idea -- make the One True Rawhide the place where we take in changes that don't break the world, while they're cobbled on in the other trees? Whether this is an extension of the "KoPeR" idea or something entirely difficult, it merits serious consideration.
I very much like the aspect of the more stable rawhide here.
Jesse Keating brought up some concerns about integration, but aren't those concerns something that people would be interested in solving? (I'm assuming those people are the wide variety of engineers working in the Fedora community who are smarter than I.)
So my plans are really funny. I plan to make rawhide more unstable more of the time, and I plan to make "rawhide" more stable more of the time. Crazy eh? How can I do this? By splitting "rawhide" in two.
Rawhide as we know it, /pub/fedora/linux/releases/development/ will remain "rawhide". We may even change the path to say rawhide, just to catch things up and well I like keeping mirrors on their toes. Rawhide will be a repository of developmental and experimental packages. Things being worked on for the future. It will /not/ be an installable tree, rather it will just be a repository of packages, to be added on to an already stable "base", eg you'd install F12, and enable rawhide to test rawhide. This will significantly lower the complaints that "rawhide isn't installable".
The second face of rawhide, will be the "pending release", that is when it comes time to feature freeze a release, we'll split it away from rawhide. We'll publish to /pub/fedora/linux/releases/test/13-pending/ or some such. THIS tree will be installable. It will be composed each night, and we'll use bodhi to manage updates to this tree. That means this tree will have it's own "updates testing" where potential freeze breaks can be tested and commented on by all, but won't risk the base tree. If testing pans out, it'll get tagged for the release, if not it'll get thrown away. This tree will spawn 13-Alpha, 13-Beta, the snapshots in between, and eventually pub/fedora/linux/releases/13.
Remember that first rawhide? Yeah, it kept going, unfrozen, leaping toward Fedora 14. You could still install 13-Alpha, or 13-pending, and enable/update to rawhide to start testing Fedora 14 stuff.
So to make this a reality, we need to ensure that whatever is in rawhide has a *>=* ENVR than anything in the other trees. So I assume that when submitting a bodhi update, bodhi would check rawhide and ensure that whatever you were about to submit to 13-pending was <= whatever was in rawhide. Otherwise we'd get into a great big mess of not being able to update to rawhide packages because whatever was in 13-pending was 'newer' than rawhide. Right?
We should have this anyway just to help upgradability between distros; bodhi should not allow a package to be added to an update if it's a "newer" ENVR than that same package in any of the "newer" distros.
Dan
On Thu, 2009-10-22 at 10:55 -0700, Dan Williams wrote:
So to make this a reality, we need to ensure that whatever is in rawhide has a *>=* ENVR than anything in the other trees. So I assume that when submitting a bodhi update, bodhi would check rawhide and ensure that whatever you were about to submit to 13-pending was <= whatever was in rawhide. Otherwise we'd get into a great big mess of not being able to update to rawhide packages because whatever was in 13-pending was 'newer' than rawhide. Right?
We should have this anyway just to help upgradability between distros; bodhi should not allow a package to be added to an update if it's a "newer" ENVR than that same package in any of the "newer" distros.
Yes, but it may happen before the bodhi stage, when we get autoqa working on post-build tests. This kind of check could happen at SCM commit time, package build time, or finally bodhi push time. Seems reasonable that we'd want to catch it as early as possible, but that does force people to work on rawhide first, then work on the pending release which may be under critical time pressure. Certainly something to discuss. Catching it at bodhi time seems too late.
On Thu, 2009-10-22 at 11:02 -0700, Jesse Keating wrote:
On Thu, 2009-10-22 at 10:55 -0700, Dan Williams wrote:
So to make this a reality, we need to ensure that whatever is in rawhide has a *>=* ENVR than anything in the other trees. So I assume that when submitting a bodhi update, bodhi would check rawhide and ensure that whatever you were about to submit to 13-pending was <= whatever was in rawhide. Otherwise we'd get into a great big mess of not being able to update to rawhide packages because whatever was in 13-pending was 'newer' than rawhide. Right?
We should have this anyway just to help upgradability between distros; bodhi should not allow a package to be added to an update if it's a "newer" ENVR than that same package in any of the "newer" distros.
Yes, but it may happen before the bodhi stage, when we get autoqa working on post-build tests. This kind of check could happen at SCM commit time, package build time, or finally bodhi push time. Seems reasonable that we'd want to catch it as early as possible, but that does force people to work on rawhide first, then work on the pending release which may be under critical time pressure. Certainly something to discuss. Catching it at bodhi time seems too late.
We could allow adding numbers after the dist tag in release for this purpose. And if the fixed package would upgrade to a newer upstream version it should not be too big hassle to build it in the rawhide tree first.
Tomas Mraz wrote:
We could allow adding numbers after the dist tag in release for this purpose.
That is already allowed, and encouraged, for branch-specific modfications, http://fedoraproject.org/wiki/Packaging/NamingGuidelines#Minor_release_bumps...
-- Rex
On Thu, 2009-10-22 at 13:39 -0500, Rex Dieter wrote:
Tomas Mraz wrote:
We could allow adding numbers after the dist tag in release for this purpose.
That is already allowed, and encouraged, for branch-specific modfications, http://fedoraproject.org/wiki/Packaging/NamingGuidelines#Minor_release_bumps...
Heh, dunno why I thought it is disallowed. Then I don't see any problem with enforcing n-v-r ordering during cvs tagging.
On Thu, 2009-10-22 at 11:02 -0700, Jesse Keating wrote:
On Thu, 2009-10-22 at 10:55 -0700, Dan Williams wrote:
So to make this a reality, we need to ensure that whatever is in rawhide has a *>=* ENVR than anything in the other trees. So I assume that when submitting a bodhi update, bodhi would check rawhide and ensure that whatever you were about to submit to 13-pending was <= whatever was in rawhide. Otherwise we'd get into a great big mess of not being able to update to rawhide packages because whatever was in 13-pending was 'newer' than rawhide. Right?
We should have this anyway just to help upgradability between distros; bodhi should not allow a package to be added to an update if it's a "newer" ENVR than that same package in any of the "newer" distros.
Yes, but it may happen before the bodhi stage, when we get autoqa working on post-build tests. This kind of check could happen at SCM commit time, package build time, or finally bodhi push time. Seems reasonable that we'd want to catch it as early as possible, but that does force people to work on rawhide first, then work on the pending release which may be under critical time pressure. Certainly something to discuss. Catching it at bodhi time seems too late.
Good point. SCM commit time (or tag time) with a CVS hook would be awesome as long as the hook was fast enough.
Dan
On Thu, 2009-10-22 at 11:33 -0700, Dan Williams wrote:
Good point. SCM commit time (or tag time) with a CVS hook would be awesome as long as the hook was fast enough.
Note, I didn't say "CVS", I said "SCM" (:
On Thu, Oct 22, 2009 at 11:02:42 -0700, Jesse Keating jkeating@redhat.com wrote:
Yes, but it may happen before the bodhi stage, when we get autoqa working on post-build tests. This kind of check could happen at SCM commit time, package build time, or finally bodhi push time. Seems reasonable that we'd want to catch it as early as possible, but that does force people to work on rawhide first, then work on the pending release which may be under critical time pressure. Certainly something to discuss. Catching it at bodhi time seems too late.
The *.fcxx.1 trick can be used to work around that if necessary. That may not be desirable though, as it seems to catch people by surprise.
On Thu, 22 Oct 2009 09:50:54 -0700, Jesse Keating jkeating@redhat.com wrote:
Hi all. It has been brought to my attention that my description of my future vision of rawhide as explained here is much clearer than previous attempts (including the current "no frozen rawhide" wiki page). []
Actually, the "no frozen rawhide" was very clear and this message just boggles the mind.
What does this accomplish? It provides a very easy release valve. Instead of closing the valve and building up pressure while we freeze, and tempting people to push things into our pending release that really don't belong, we'll provide them a normal, never ending release of pressure, called rawhide. []
Great, you made life of lazy developers easy. What about users of Rawhide? I had Rawide installed on my main desktop since RHL 6. What am I supposed to do now? What tree to run? Your proposal has no guidelines for me (unlike the proposal to stop freezing Rawhide, which IMHO had a lot of merit).
Note, I'm not interested in testing things "sometimes". If I did, I would've been using RHEL WS or something. I want an always-useable development tip tree.
The problem here is that Rawhide was a useful distribution for years. We were Gentoo before Gentoo, minus meaningless recompiles. And now?
-- Pete
On Thu, 2009-10-22 at 14:47 -0600, Pete Zaitcev wrote:
On Thu, 22 Oct 2009 09:50:54 -0700, Jesse Keating jkeating@redhat.com wrote:
Hi all. It has been brought to my attention that my description of my future vision of rawhide as explained here is much clearer than previous attempts (including the current "no frozen rawhide" wiki page). []
Actually, the "no frozen rawhide" was very clear and this message just boggles the mind.
What does this accomplish? It provides a very easy release valve. Instead of closing the valve and building up pressure while we freeze, and tempting people to push things into our pending release that really don't belong, we'll provide them a normal, never ending release of pressure, called rawhide. []
Great, you made life of lazy developers easy. What about users of Rawhide? I had Rawide installed on my main desktop since RHL 6. What am I supposed to do now? What tree to run? Your proposal has no guidelines for me (unlike the proposal to stop freezing Rawhide, which IMHO had a lot of merit).
Note, I'm not interested in testing things "sometimes". If I did, I would've been using RHEL WS or something. I want an always-useable development tip tree.
The problem here is that Rawhide was a useful distribution for years. We were Gentoo before Gentoo, minus meaningless recompiles. And now?
-- Pete
I wonder where your confusion comes from. With this rewording of the proposal, the proposal doesn't change. Rawhide the path on the mirror (pub/fedora/linux/releases/development/) never freezes, never stops. It's always developmental and testing packages. It is always there. It never freezes. You can turn on that repo and just run with it from now until you decide you don't want to use computers anymore.
The proposal, both wordings, say that we'll stop slowing down rawhide, stop trying to use that path as a testing/polish point for a release. We'll do that somewhere else and let rawhide keep flowing.
So you can continue to run rawhide all you want. Your entry point to rawhide may change slightly, you may have to start with the current Fedora release or the current testing release for the next Fedora, and then upgrade to the rawhide package set, but once you've done that you just stick on rawhide and never look back.
Le Jeu 22 octobre 2009 23:20, Jesse Keating a écrit :
So you can continue to run rawhide all you want. Your entry point to rawhide may change slightly, you may have to start with the current Fedora release or the current testing release for the next Fedora, and then upgrade to the rawhide package set, but once you've done that you just stick on rawhide and never look back.
IMHO you should be honest and call your F-next rawhide, and find some other name for the "experimental stuff" sandbox. Because your F-next is effectively what rawhide has been since before Fedora, and your rawhide is what people who do not care about releng have been trying to turn rawhide into in the past years. Well they can get their no-rules playground but I don't see why we should also give them the rawhide name. It has a fine history, much better than the eating babies ogre caricature.
On Thu, 22 Oct 2009 14:20:13 -0700, Jesse Keating jkeating@redhat.com wrote:
I wonder where your confusion comes from. With this rewording of the proposal, the proposal doesn't change. []
So you can continue to run rawhide all you want. Your entry point to rawhide may change slightly, you may have to start with the current Fedora release or the current testing release for the next Fedora, and then upgrade to the rawhide package set, but once you've done that you just stick on rawhide and never look back.
That works for me, thanks a lot.
-- Pete
On Thu, 2009-10-22 at 09:50 -0700, Jesse Keating wrote:
Hi all. It has been brought to my attention that my description of my future vision of rawhide as explained here is much clearer than previous attempts (including the current "no frozen rawhide" wiki page). So I felt it prudent to forward it along to the devel list for more eyes to look upon it and comment if desired.
I have two particular nits with it. One, it's pretty unwieldy, especially for part time maintainers (thinking how many hoops we'll have to jump through just to keep our packages up to date). Having to jump through the Bodhi hoops every time we just want to put a trivial update into a release that won't be coming out for five months feels like a pain. I'd worry about a lot of stuff going stale and smelly in the middle of the Bodhi process somewhere as maintainers lose track of where the hell they're up to with the four releases they have to cope with (the last two already-released releases, the upcoming release, and "rawhide").
Two, it makes testing things a bit more complex. Those of us who like to test upcoming stuff in real use - i.e. on our main machines - will have to choose whether to test "rawhide", in which case we'll have more pain to deal with ourselves and won't be contributing as much to testing of the next stable release, or test the next stable release, in which case we aren't helping maintainers by making sure the stuff they're putting in "rawhide" isn't totally broken.
But overall, the positives could certainly outweigh the negatives. Just thought I'd flag up the two major concerns I have in case anything could be done about them.
On Thu, 2009-10-22 at 14:18 -0700, Adam Williamson wrote:
I have two particular nits with it. One, it's pretty unwieldy, especially for part time maintainers (thinking how many hoops we'll have to jump through just to keep our packages up to date). Having to jump through the Bodhi hoops every time we just want to put a trivial update into a release that won't be coming out for five months feels like a pain.
Who said anything about 5 months. 3 months max, which is just slightly longer than the amount of time they spend now doing Trac tickets to get something in.
I'd worry about a lot of stuff going stale and smelly in the middle of the Bodhi process somewhere as maintainers lose track of where the hell they're up to with the four releases they have to cope with (the last two already-released releases, the upcoming release, and "rawhide").
"rawhide" would never use bodhi. If you build in devel/ it shows up in rawhide the next day. End of story. Bodhi would only be used for the pending and already done releases.
Two, it makes testing things a bit more complex. Those of us who like to test upcoming stuff in real use - i.e. on our main machines - will have to choose whether to test "rawhide", in which case we'll have more pain to deal with ourselves and won't be contributing as much to testing of the next stable release, or test the next stable release, in which case we aren't helping maintainers by making sure the stuff they're putting in "rawhide" isn't totally broken.
For people like you, you'd just keep jumping when we branch off the next release. Like it or not, this is happening /already/. dist-f13 already has 953 packages with updates, and it isn't published /anywhere/ for /anybody/ to test it. We can only make that better, not worse.
But overall, the positives could certainly outweigh the negatives. Just thought I'd flag up the two major concerns I have in case anything could be done about them.
I think this is an awesome idea, and yes I think this "version" of the verbage is more clear. Kudos to Jesse (and all those involved in the development of the idea of the "split rawhide") and I hope to see this come to fruition.
-Adam
On Thu, 22 Oct 2009, Adam Miller wrote:
I think this is an awesome idea, and yes I think this "version" of the verbage is more clear. Kudos to Jesse (and all those involved in the development of the idea of the "split rawhide") and I hope to see this come to fruition.
I agree, how can I help?
-Mike
I would also like to jump on the help train if there is anything I am able to lend a hand with.
-Adam (From Android)
On Oct 22, 2009 9:05 PM, "Mike McGrath" mmcgrath@redhat.com wrote:
On Thu, 22 Oct 2009, Adam Miller wrote: > I think this is an awesome idea, and yes I think this "ve... I agree, how can I help?
-Mike
-- fedora-devel-list mailing list fedora-devel-list@redhat.com https://www.redhat.com/mailman/list...
Dne 23.10.2009 02:18, Adam Miller napsal(a):
I think this is an awesome idea, and yes I think this "version" of the verbage is more clear. Kudos to Jesse (and all those involved in the development of the idea of the "split rawhide") and I hope to see this come to fruition.
Just wanted to add my +1 and this is as good place as any other.
Matěj
On Fri, Oct 23, 2009 at 8:24 AM, Matěj Cepl mcepl@redhat.com wrote:
Just wanted to add my +1 and this is as good place as any other.
+1
On Thu, Oct 22, 2009 at 14:18:23 -0700, Adam Williamson awilliam@redhat.com wrote:
Two, it makes testing things a bit more complex. Those of us who like to test upcoming stuff in real use - i.e. on our main machines - will have to choose whether to test "rawhide", in which case we'll have more pain to deal with ourselves and won't be contributing as much to testing of the next stable release, or test the next stable release, in which case we aren't helping maintainers by making sure the stuff they're putting in "rawhide" isn't totally broken.
The impression I got (which might be wrong), is that it was expected that people would test specific packages from rawhide and not be expected to be running it all at once. Porbably the best equvialent to current rawhide would be enabling updates-testing for the pending release. People that needed some that wouldn't break unexpectedly, but still wanted to try out the new stuff could run the pending release without updates-testing.
Bruno Wolff III wrote:
The impression I got (which might be wrong), is that it was expected that people would test specific packages from rawhide and not be expected to be running it all at once.
That just doesn't work. The network of dependencies and reverse dependencies generally ends up dragging in half of the distro after the first core library soname bump (e.g. OpenSSL, OpenLDAP or the like).
Kevin Kofler
On Thu, Oct 22, 2009 at 4:50 PM, Jesse Keating jkeating@redhat.com wrote:
Rawhide as we know it, /pub/fedora/linux/releases/development/ will remain "rawhide". We may even change the path to say rawhide, just to catch things up and well I like keeping mirrors on their toes. Rawhide will be a repository of developmental and experimental packages. Things being worked on for the future. It will /not/ be an installable tree, rather it will just be a repository of packages, to be added on to an already stable "base", eg you'd install F12, and enable rawhide to test rawhide. This will significantly lower the complaints that "rawhide isn't installable".
So as I understand it there are a number of reasons why rawhide might not be installable, but broadly they fall into two major categories:
* Anaconda * Critpath packages - Dependency/rebuild issues - Bugs in %posts (like the user/group one we ran into with dbus) - Core bugs (graphics drivers)
It seems like we're basically just skipping Anaconda, since you won't be able to yum if there are depsolving issues (ok, modulo --skip-broken), and for the latter two you don't end up with a "working" system.
Let me do a counter-proposal:
We simply do not let showstopper regressions in the critpath stay in rawhide. If something in critpath has a showstopper, it halts all further commits to the entire critpath until it's resolved (either fixed, or reverted). The definition of "showstopper" might be "AutoQA fails". And since AutoQA will have been doing some basic smoketesting of the installer, we have to be producing installer images as a side effect.
On Fri, 2009-10-23 at 20:56 +0000, Colin Walters wrote:
On Thu, Oct 22, 2009 at 4:50 PM, Jesse Keating jkeating@redhat.com wrote:
Rawhide as we know it, /pub/fedora/linux/releases/development/ will remain "rawhide". We may even change the path to say rawhide, just to catch things up and well I like keeping mirrors on their toes. Rawhide will be a repository of developmental and experimental packages. Things being worked on for the future. It will /not/ be an installable tree, rather it will just be a repository of packages, to be added on to an already stable "base", eg you'd install F12, and enable rawhide to test rawhide. This will significantly lower the complaints that "rawhide isn't installable".
So as I understand it there are a number of reasons why rawhide might not be installable, but broadly they fall into two major categories:
- Anaconda
- Critpath packages
- Dependency/rebuild issues
- Bugs in %posts (like the user/group one we ran into with dbus)
- Core bugs (graphics drivers)
It seems like we're basically just skipping Anaconda, since you won't be able to yum if there are depsolving issues (ok, modulo --skip-broken), and for the latter two you don't end up with a "working" system.
Let me do a counter-proposal:
We simply do not let showstopper regressions in the critpath stay in rawhide. If something in critpath has a showstopper, it halts all further commits to the entire critpath until it's resolved (either fixed, or reverted). The definition of "showstopper" might be "AutoQA fails". And since AutoQA will have been doing some basic smoketesting of the installer, we have to be producing installer images as a side effect.
I... don't see how this helps, other than piss off the rest of the crit-path maintainers while one thing is broken.
AutoQA will be running at some point, and it can be doing the qa /before/ things get tagged for rawhide, so if you break deps with your build, it doesn't get in, unless you force it and then you face the wrath of releng/qa. To catch core bugs, we'll need a bit more advanced autoqa, doing more than just repo level testing but doing actual package testing. That will grow over time and again can be done pre-tag.
Your counter proposal also does nothing to help the dual or sometimes triple role we try to put on "rawhide" the path.
On Fri, Oct 23, 2009 at 9:11 PM, Jesse Keating jkeating@redhat.com wrote:
AutoQA will be running at some point, and it can be doing the qa /before/ things get tagged for rawhide,
Oh, I didn't realize there would be a distinction between "built in koji" and "rawhide" now. If that's the case, than this sounds fine to me! The point is basically that we need some sort of stable, defined baseline for what you get when you try to install "rawhide". Testing before tagging sounds good.
On Fri, 2009-10-23 at 21:15 +0000, Colin Walters wrote:
Oh, I didn't realize there would be a distinction between "built in koji" and "rawhide" now. If that's the case, than this sounds fine to me! The point is basically that we need some sort of stable, defined baseline for what you get when you try to install "rawhide". Testing before tagging sounds good.
Yeah, there are multiple prongs to our attack to make Fedora development better. In this thread I was mostly talking about what types of changes go into the repos and which repos get produced each night. Autoqa has other plans to help, one of which being testing before tagging.