Hi,
I want to update the SIG on various news:
* ghc-7.4.2 and haskell-platform-2012.4
I have started work locally testing updating rawhide to ghc-7.4.2 and haskell-platform-2012.4 RC2. If people want I can upload my local ghc-7.4.2 and haskell-platform builds for testing.
Overall the results so far are fairly good. Only a few remaining packages have problems:
- Agda (constructor issue) - happstack-server (blaze-html or new deps) - hledger-lib-0.19 needs shakespeare-text (review pending) - xss-sanitize (css-text not imported yet!)
but they should be fixable by F19 time.
I plan to build ghc-7.4.2 for Rawhide within the next week.
* Updating packages to latest versions
We have a lot of packages that need a version refresh. Now is a great time to bump your packages in git master. Building for rawhide is also ok but not really necessary right now since everything will need rebuilding shortly soon anyway. I really need your help with this, but if you don't have time to do it I will start touching packages generally in the coming week or so.
* cabal-rpm update
I have been using cabal-rpm to refresh a few packages by hand and I am going to release 0.6.5 soon with some smaller tweaks and improvements: the biggest visible change being no longer using version-ranged BRs since I think this has too much maintenance overhead at least until cabal-rpm can do updating of packages.
* Haskell Platform and ghc versions
I am wondering how closely we should follow Haskell Platform going forward since the releases are not well synced with Fedora's schedule and upstream so far refused to shift its schedule...
As such I quite like following HP - it is gives us a certain stability and certainty of base versions, but also understand that some people want to have the latest versions available and live on the bleeding edge. Personally I am leaning towards doing more backporting - I am thinking to backport ghc-7.4.2 and haskell-platform-2012.4 to F18 for example (after it has gone to rawhide). If there is demand we could also backport ghc-7.4.1 to F17 perhaps or I could make a yum repo for it from the F18 packages. Ubuntu-12.10 just recently shipped with ghc-7.4.2.
What do people think? How important is it to have the official haskell-platform in Fedora?
Another side issue is that I somewhat cavalierly made haskell-platform into a monolithic package in F18, thinking this would decrease its update/build time. This unfortunately means if we weren't going to follow HP closely we would probably need to re-review all the HP packages again... :-/
Jens
I personally think Platform is a bit incompatible with the Fedora philosophy of "features" and "first". I don't think we need to be the *bleeding* edge, but at least the *cutting* edge? I think people who install and use [the latest] Fedora have already made a decision that they want to experiment with the latest in software development, and usually have the skill to do so. Those who just want something that works with minimum fuzz and aren't too interested in trying out the latest GHC features are likely to be running something other than Fedora, I suspect.
It seems Fedora releases are usually around two or three months behind GHC releases, which is also about the time it takes for Hackage to stabilize to a new GHC release. That's what I mean about cutting vs. bleeding edge.
Who wants to be stuck with GHC 7.4 for at least the next seven or eight months? Who wants to have GHC 7.6 within the next month or two, at the cost of not having the precise versions of packages specified in the latest Platform?
I know my post on how to get GHC 7.4 in F17 has been very popular, and I'm hoping we can avoid repeating this situation for GHC 7.6 and F18.
Just my two cents.
-- donri
On Mon, Oct 29, 2012 at 11:16 AM, Jens Petersen petersen@redhat.com wrote:
Hi,
I want to update the SIG on various news:
- ghc-7.4.2 and haskell-platform-2012.4
I have started work locally testing updating rawhide to ghc-7.4.2 and haskell-platform-2012.4 RC2. If people want I can upload my local ghc-7.4.2 and haskell-platform builds for testing.
Overall the results so far are fairly good. Only a few remaining packages have problems:
- Agda (constructor issue)
- happstack-server (blaze-html or new deps)
- hledger-lib-0.19 needs shakespeare-text (review pending)
- xss-sanitize (css-text not imported yet!)
but they should be fixable by F19 time.
I plan to build ghc-7.4.2 for Rawhide within the next week.
- Updating packages to latest versions
We have a lot of packages that need a version refresh. Now is a great time to bump your packages in git master. Building for rawhide is also ok but not really necessary right now since everything will need rebuilding shortly soon anyway. I really need your help with this, but if you don't have time to do it I will start touching packages generally in the coming week or so.
- cabal-rpm update
I have been using cabal-rpm to refresh a few packages by hand and I am going to release 0.6.5 soon with some smaller tweaks and improvements: the biggest visible change being no longer using version-ranged BRs since I think this has too much maintenance overhead at least until cabal-rpm can do updating of packages.
- Haskell Platform and ghc versions
I am wondering how closely we should follow Haskell Platform going forward since the releases are not well synced with Fedora's schedule and upstream so far refused to shift its schedule...
As such I quite like following HP - it is gives us a certain stability and certainty of base versions, but also understand that some people want to have the latest versions available and live on the bleeding edge. Personally I am leaning towards doing more backporting - I am thinking to backport ghc-7.4.2 and haskell-platform-2012.4 to F18 for example (after it has gone to rawhide). If there is demand we could also backport ghc-7.4.1 to F17 perhaps or I could make a yum repo for it from the F18 packages. Ubuntu-12.10 just recently shipped with ghc-7.4.2.
What do people think? How important is it to have the official haskell-platform in Fedora?
Another side issue is that I somewhat cavalierly made haskell-platform into a monolithic package in F18, thinking this would decrease its update/build time. This unfortunately means if we weren't going to follow HP closely we would probably need to re-review all the HP packages again... :-/
Jens _______________________________________________ haskell mailing list haskell@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/haskell
On Mon, Oct 29, 2012 at 10:49 AM, dag.odenhall@gmail.com dag.odenhall@gmail.com wrote:
Who wants to be stuck with GHC 7.4 for at least the next seven or eight months? Who wants to have GHC 7.6 within the next month or two, at the cost of not having the precise versions of packages specified in the latest Platform?
I must admit, I would prefer to have GHC 7.6 sooner rather than later.
Cheers,
Miles
Who wants to be stuck with GHC 7.4 for at least the next seven or eight months? Who wants to have GHC 7.6 within the next month or two, at the cost of not having the precise versions of packages specified in the latest Platform?
I must admit, I would prefer to have GHC 7.6 sooner rather than later.
Ok. As a step towards this goal I think backporting offers quite a big improvement in this area and will free us largely from the current Haskell Platform schedule skew at least, since we can update to newest HP as soon as it is available. But like for Windows and Mac users it will be another 6+ months before we see ghc-7.6+ in Fedora (ie HP 2013.2).
Has anyone tried to rebuild all of our packages with 7.6.1 yet? I just finished testing 7.4.2 as I mentioned. I could imagine there will still be quite a few more problems with 7.6 remaining. It is not without reason that HP decided to stick with 7.4.2 at this time for the coming 2012.4 release.
Anyway help and input is certainly appreciated and welcome. If I was working full-time on Fedora Haskell I could really do a lot more...
Thanks, Jens
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 10/29/2012 05:16 PM, Jens Petersen wrote:
Hi,
I want to update the SIG on various news:
- ghc-7.4.2 and haskell-platform-2012.4
...
- hledger-lib-0.19 needs shakespeare-text (review pending)
The packager in question has more or less disappeared over the past couple of months; given that shakespeare-text is a dependency of yesod (which he has also submitted for review), which I'm keenly interested in, should we do something about it? I'd be happy to re-submit the packages making up the Yesod stack for review over the next few days.
What do people think? How important is it to have the official haskell-platform in Fedora?
I'm quite happy with the backport approach, sort of like how KDE is maintained in Fedora. And that ties in quite well with keeping haskell-platform, right? At least then when newer versions of GHC make it to older Fedora releases, it'd be less of a mess rebuilding all the components.
Regards,
- -- Michel Alexandre Salim Fedora Project Contributor: http://fedoraproject.org/
Email: salimma@fedoraproject.org | GPG key ID: A36A937A Jabber: hircus@jabber.ccc.de | IRC: hircus@irc.freenode.net
() ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
given that shakespeare-text is a dependency of yesod (which he has also submitted for review), which I'm keenly interested in, should we do something about it? I'd be happy to re-submit the packages making up the Yesod stack for review over the next few days.
Yes, Ben actually mentioned to me that you are keen to get Yesod in. It is something that we have wanted for quite a while but found to be a bit of a moving target to keep up with with our limited resources. Submitting all the packages is not that hard - getting them reviewed more so... So in that sense might be better if someone like you could help to review them. :-) Anyway how about doing a Yesod Feature for F19? It is certainly something I would also like to see happen and I will try to help.
What do people think? How important is it to have the official haskell-platform in Fedora?
I'm quite happy with the backport approach, sort of like how KDE is maintained in Fedora. And that ties in quite well with keeping haskell-platform, right? At least then when newer versions of GHC make it to older Fedora releases, it'd be less of a mess rebuilding all the components.
Right perhaps we should try to keep all the releases in sync. Though it still seems a somewhat of daunting undertaking to me. But basically thanks to Haskell once rawhide is good then mass backporting to older releases should just work generally. It might be a bit too much churn for EPEL though perhaps.
Jens
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 30/10/2012 16:14, Jens Petersen wrote:
given that shakespeare-text is a dependency of yesod (which he has also submitted for review), which I'm keenly interested in, should we do something about it? I'd be happy to re-submit the packages making up the Yesod stack for review over the next few days.
Yes, Ben actually mentioned to me that you are keen to get Yesod in. It is something that we have wanted for quite a while but found to be a bit of a moving target to keep up with with our limited resources. Submitting all the packages is not that hard - getting them reviewed more so... So in that sense might be better if someone like you could help to review them. :-) Anyway how about doing a Yesod Feature for F19? It is certainly something I would also like to see happen and I will try to help.
Sounds like a good idea. I'll start a feature page in the next couple of days, and we can probably divide the work up, taking turns packaging and reviewing.
Should I list you as a feature co-owner? And maybe Ben too once he gets back.
Also, what to do with review requests that have bit-rotted?
Thanks,
- -- Michel Alexandre Salim Fedora Project Contributor: http://fedoraproject.org/
Email: salimma@fedoraproject.org | GPG key ID: A36A937A Jabber: hircus@jabber.ccc.de | IRC: hircus@irc.freenode.net
() ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
Hi Michel,
I'll start a feature page in the next couple of days, and we can probably divide the work up, taking turns packaging and reviewing.
That would be great!
Should I list you as a feature co-owner? And maybe Ben too once he gets back.
Sounds good - yes please. Probably we could comaintain the packages.
Also, what to do with review requests that have bit-rotted?
Probably easiest just to close them as duplicates of the new review requests.
Thanks, Jens
Hi,
Fedora 18 is now frozen for release - so a good time now to start thinking about features for Fedora 19.
Michel wrote:
[Yesod] is something that we have wanted for quite a while but found to be a bit of a moving target to keep up with with our limited resources. Submitting all the packages is not that hard - getting them reviewed more so... So in that sense might be better if someone like you could help to review them. :-) Anyway how about doing a Yesod Feature for F19? It is certainly something I would also like to see happen and I will try to help.
Sounds like a good idea. I'll start a feature page in the next couple of days, and we can probably divide the work up, taking turns packaging and reviewing.
Ok, I went ahead and created a draft Feature page:
https://fedoraproject.org/wiki/Features/YesodWebFramework
I think it is a reasonable start but still a bit terse and lacking some details so feel free to edit and improve it. The submission deadline is end of January but I don't see why we should not be able to submit it earlier next month after scoping a little more.
So far I added Ben, Michel, and Narasim as co-owners.
Maybe the largest missing detail is the number of packages still needed and the exact list of them: if someone has prepared that then please add it as a link to the Feature page.
I guess we have a tracking bug to measure our progress too.
Also, what to do with review requests that have bit-rotted?
Maybe easier just close them now if they haven't been refreshed.
Thanks, Jens
On Dec 11, 2012 4:34 PM, "Jens Petersen" petersen@redhat.com wrote:
Hi,
Fedora 18 is now frozen for release - so a good time now to start thinking about features for Fedora 19.
...
Ok, I went ahead and created a draft Feature page:
https://fedoraproject.org/wiki/Features/YesodWebFramework
I think it is a reasonable start but still a bit terse and lacking some details so feel free to edit and improve it. The submission deadline is end of January but I don't see why we should not be able to submit it earlier next month after scoping a little more.
Ah, apologies, I've had a draft sitting on my laptop for a while that I have not found the time to spruce up for publication.
Will merge that into the existing draft.
So far I added Ben, Michel, and Narasim as co-owners.
Maybe the largest missing detail is the number of packages still needed and the exact list of them: if someone has prepared that then please add it as a link to the Feature page.
That, and packages that need updating.
Thanks,
- Michel (boy, Android is a painful device to use with mailing list communications)
Ok, I went ahead and created a draft Feature page: https://fedoraproject.org/wiki/Features/YesodWebFramework
Ah, apologies, I've had a draft sitting on my laptop for a while that I have not found the time to spruce up for publication. Will merge that into the existing draft.
That would be great, thanks!
Maybe the largest missing detail is the number of packages still needed and the exact list of them: if someone has prepared that then please add it as a link to the Feature page.
That, and packages that need updating.
Yes, I think we need some status wiki page to track all the dependencies, etc.
Jens
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 25/12/12 12:34, Jens Petersen wrote:
Ok, I went ahead and created a draft Feature page: https://fedoraproject.org/wiki/Features/YesodWebFramework
Ah, apologies, I've had a draft sitting on my laptop for a while that I have not found the time to spruce up for publication. Will merge that into the existing draft.
That would be great, thanks!
Maybe the largest missing detail is the number of packages still needed and the exact list of them: if someone has prepared that then please add it as a link to the Feature page.
That, and packages that need updating.
Yes, I think we need some status wiki page to track all the dependencies, etc.
I did a preliminary script -- it works, but I'm not sure exactly how to check, for a given Hackage entry, what the corresponding Fedora package is, so there are a few false positives (I did the simple 'check if a ghc-...' binary RPM exist')
Here's what it finds (in ascending order) -- 61 in total. Any suggestion for improving the detection of known packages would be very welcome; I'll work on generating MediaWiki-formatted dependency maps once that's sorted.
[u'ReadArgs', u'Win32', u'asn1-data', u'attoparsec-conduit', u'authenticate', u'base64-conduit', u'basic-prelude', u'certificate', u'classy-prelude', u'classy-prelude-conduit', u'clientsession', u'cookie', 'cprng-aes', u'cpu', u'crypto-api', u'crypto-pubkey-types', u'crypto-random-api', u'cryptocipher', u'email-validate', u'entropy', u'file-embed', u'fsnotify', u'http-conduit', u'http-reverse-proxy', u'largeword', u'mime-mail', u'monad-logger', u'network-bytestring', u'optparse-applicative', u'path-pieces', u'pem', u'persistent', u'persistent-template', u'pool-conduit', u'project-template', u'pureMD5', u'pwstore-fast', u'resource-pool', u'shakespeare-css', u'shakespeare-i18n', u'shakespeare-js', u'shakespeare-text', u'silently', u'skein', u'socks', u'system-fileio', u'system-filepath', u'tls', u'tls-extra', u'word8', u'xml-conduit', u'xml-types', u'yaml', 'yesod', u'yesod-auth', u'yesod-core', u'yesod-default', u'yesod-form', u'yesod-json', u'yesod-persistent', u'yesod-routes']
Thanks,
- -- Michel Alexandre Salim Fedora Project Contributor: http://fedoraproject.org/
Email: salimma@fedoraproject.org | GPG key ID: A36A937A Jabber: hircus@jabber.ccc.de | IRC: hircus@irc.freenode.net
() ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
I went ahead and finally submitted the Feature yesterday just before the submission deadline. :)
I did a preliminary script -- it works, but I'm not sure exactly how to check, for a given Hackage entry, what the corresponding Fedora package is, so there are a few false positives (I did the simple 'check if a ghc-...' binary RPM exist')
Okay thanks sounds like a good start - sorry I didn't have time to look at it yet but I will soon.
Here's what it finds (in ascending order) -- 61 in total.
Okay that is quite a lot but if we can "fast-track" them I think it should be possible - at least first goal should be to complete yesod-core. cabal-rpm should also help to make it easier now to generate packages since it now generates BR dependencies.
Jens
I went ahead and finally submitted the Feature yesterday just before the submission deadline. :)
I did a preliminary script -- it works, but I'm not sure exactly how to check, for a given Hackage entry, what the corresponding Fedora package is, so there are a few false positives (I did the simple 'check if a ghc-...' binary RPM exist')
Okay thanks sounds like a good start - sorry I didn't have time to look at it yet but I will soon.
Here's what it finds (in ascending order) -- 61 in total.
We have passed F19 Feature Freeze date and not made much progress yet on Yesod. I suggest we can move this feature to Fedora 20 - though it doesn't stop us from continuing to work on adding required packages to Fedora now - even we could still do it if we had the time and manpower, but with the current slow progress it seems unlikely we can complete it for F19. Yesod-core might be realistic perhaps.
http://fedoraproject.org/wiki/Releases/19/Schedule
There are various substacks that would certainly be great to work on and splitting up the work in that way also makes it more manageable and realistic: eg client-session, http-conduit (tls), yesod-core, etc but there is also a longish list of individual deps that will already build or ones with just a single dep missing, etc. Let's get started and keep pushing on this.
Thanks, Jens
Hi,
I made some more progress towards the big rawhide update to ghc-7.4.2 and new HP:
- Agda (constructor issue)
fixed
- happstack-server (blaze-html or new deps)
7.0 patched until we can upgrade to 7.1.
- hledger-lib-0.19 needs shakespeare-text (review pending)
For now I got 0.17 to build.
- xss-sanitize (css-text not imported yet!)
Imported and builds
I plan to build ghc-7.4.2 for Rawhide within the next week.
I will be travelling for the next days and nearly all the packages without a version bump still need to be release bumped and ideally refreshed for cabal-rpm. (Actually I have scripting with a generic patch for that, so while probably it is quicker for me to do it, help is appreciated and it would also help you understand better the new cabal-rpm template changes.) Anyway please do not build any of the bumped packages yet until ghc-7.4.2, etc are in rawhide: I will send a heads-up at the end of next week or when I start building to let you know. Ask me if you are interested in testing and I can explain how to rebuild everything or upload my current x86_64 packages for F18 to fedorapeople.
We have a lot of packages that need a version refresh. Now is a great time to bump your packages in git master. Building for rawhide is also ok but not really necessary right now since everything will need rebuilding shortly soon anyway. I really need your help with this, but if you don't have time to do it I will start touching packages generally in the coming week or so.
This is basically all done now by me - I don't think any more packages can be bumped right now without needing additional new deps or causing breakage.
* cabal-rpm
I find typing the command "cabal-rpm" a bit long so I think I will probably rename the executable to cblrpm in the next release.
Cheers, Jens
I will send a heads-up at the end of next week or when I start building to let you know.
I finished bumping packages in git master branches over the weekend and building for ghc-7.4.2 is now well underway.
As I write, currently about 70 packages still need building. I think it all should be largely finished within the next day or two. I will send out a sign-off mail when that happens.
Jens
Hi,
I think it all should be largely finished within the next day or two. I will send out a sign-off mail when that happens.
Nearly everything in Rawhide is rebuilt now (actually since yesterday). Probably the only exception is ghc-wai-extra-1.3 which needs 3 new dependencies, but I wanted to get wai-1.3 into rawhide now anyway...
Next major milestone will be to backport this work to F18.
The rawhide ghc-7.4.2 update "freeze" is over and normal development can start again. :)
Thanks, Jens
Hi,
Nearly everything in Rawhide is rebuilt now (actually since yesterday). Probably the only exception is ghc-wai-extra-1.3 which needs 3 new dependencies, but I wanted to get wai-1.3 into rawhide now anyway...
Next major milestone will be to backport this work to F18.
Ok a few other concerns:
1. I want to finally get the revised Haskell Packaging draft back before FPC asap. I will update it for cabal-rpm and post a RFC here.
2. I am starting to think that my monolithic haskell-platform might be kind of a mistake. So I may submits platform packages for re-review for F19, sorry. The packages should not have changed hugely since they were deprecated for f18 so I think the re-reviews should be reasonably straightforward checks. (Anyway that accounts for the broken deps for primitive, split and vector which are now in latest HP.)
3. Need to make rebuild-koji.sh parallel - I think it should be easy to do - just need it to repo-wait for latest NVRs from pkg git rather than the latest koji builds.
The rebuilding this time took about 3 days (also partly due to me not testing all the bumped pkgs) - but parallelizing it could probably be done in one day which is pretty good (of course the largest bottleneck is still mock).
4. The Yesod feature proposal for F19 - if noone else starts soon I will start to draft something next week hopefully.
I think 1 should be done before 2 and 4...
Thanks, Jens
haskell@lists.fedoraproject.org