Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 6 Next »

 alanc: btw, one additional word of warning - unless you are absolutely positively sure that everyone has version 141 or later of the pkg gate bits installed, even when they are starting on 134 and upgrading to your bits, you *really really really* want the b146 version of SUNWxscreensaver.spec from JDS
[18:06] alanc: or to manually delete the "" in the description in the older specs
[18:06] richlowe: You're going to end up using the importer anyway, so hacking it by hand is just as easy.
[18:06] Alasdairrr: *makes notes*
[18:06] alanc: the nested quoting in the description breaks IPS in so many many different ways
[18:06] richlowe: all that's fixed now, isn't it?
[18:06] richlowe: I only realize when I've bothered to check upgrade from 134
[18:07] alanc: except IPS thinks it's all fixed now, so trying to image-update from older builds feeds them descriptions they can't hack
[18:07] richlowe: Right, that's what I meant. :\
[18:08] blu joined the chat room.
[18:08] alanc: I think I've also seen errors from using mismatched bits with AI, and the repo-dvd generation for 144 failed because RE ran it on a machine with an out-of-date set of IPS bits
[18:08] richlowe: Hmmm, I have a bunch of aclocal hackery in my jds workspace.
[18:09] richlowe: I'm not sure if that's something y'all would end up needing, I think it related to cheating my way out of having to upgrade SFW before starting the JDS build.
[18:09] richlowe: Alasdairrr: Also, another X note.  You need autopoint, which you won't have until you build a newer SFW
[18:09] alanc: I can't claim to have been clever enough to have planned this as a smokescreen so people didn't notice the disruption from the X refactoring in the same build, but it worked
[18:09] richlowe: so put /opt/dtbld bin into your $PATH, to use the JDS CBE bits that do have autopoint


[18:46] richlowe: SUNWbrltty is one that needed to be forced to use /usr/bin/ld, SUNWgnome-themes has a dubious dependency on SUNWneutral-plus-cursors, which I think is now obsolete, SUNWprint-monitor depends upon SUNWpc and SUNWps in a way that never seemed to be satisfiable, and you need to hack whichever packages you've not actually built out of SUNWzz-gnome-l10n.  And the aclocal hackery to get GTKDOC* macros expanded, which as I said, may not
[18:46] richlowe: be needed now.
[18:46] richlowe: there was another package that needed the LD hack, but I didn't, apparently, change the spec, so I dunno which it was.
[18:46] richlowe: you'll recognize it by it being the package that makes GNU ld crash...


[22:03] Alasdairrr: http://openindiana.pastebin.org/436730
[22:03] Alasdairrr: Odd errors
[22:04] richlowe: Expected errors
[22:05] richlowe: The SVR4 package build doesn't work without the closed sources, because the package boundaries weren't ever set up for that.
[22:05] richlowe: also, why would you build anything old enough that it still had SVR4 packages?
[22:06] Alasdairrr: Hmm
[22:06] jamesog: Does b134 still have SVR4 packages then?
[22:06] Alasdairrr: The closed sources were untarred as per documentation
[22:06] richlowe: Yes.
[22:07] richlowe: No, they're the closed binaries.
[22:07] Alasdairrr: ohh
[22:07] Alasdairrr: i see
[22:07] richlowe: And yeah, ON switched to IPS packaging in build 136
[22:07] jamesog: Ah, lovely
[22:07] richlowe: there are some binaries that can't be redistrbuted, so aren't even closed-bins.
[22:07] richlowe: and the SVR4 packages comingled them.

Alasdairrr: richlowe: Have you built slim_install?
[16:28] aszeszo: I'm following Rich's notes so the answer would be yes ;)
[16:29] Alasdairrr: haha
[16:29] Alasdairrr: :D
[16:31] richlowe: yeah, I'm currently running 144 of most things.
[16:31] richlowe: didn't bother with g11n or man
[16:31] richlowe: because man would actually be a downgrade
[16:33] richlowe: like I said, the real pain is JDS.
[16:33] richlowe: and the fact that once you get here, you can't upload bits piecemeal anymore, because of the inter-dependence.
[16:34] Alasdairrr: yeah
[16:34] richlowe: the results of what I did are maybe 2.5G/build
[16:34] richlowe: (x86 only)
[16:35] Alasdairrr: For building a clone livecd, presumably we'd need: SFW, X, JDS - is there anything else we'd need?
[16:35] Alasdairrr: Such as the man pages or docs, network storage
[16:37] chrisridd: git-svn has just  taken 20 minutes to import what hgsubversion has taken 4 hours to import. But will the results be any use?
[16:37] Alasdairrr: Do you now have to go from git to hg?
[16:38] richlowe: hg convert would probably have some success.
[16:38] richlowe: Alasdairrr: you can build with a pretty wide mix of stuff.
[16:38] richlowe: you're not going to get a completely self-built CD without more work.
[16:38] richlowe: (I'm not sure how much, but the list of packages is in the slim_install package)
[16:39] Alasdairrr: I don't suppose you could clarify 'more work' - what things are you thinking of?
[16:39] richlowe: Specifically, nspr
[16:39] chrisridd: Mozilla...
[16:39] richlowe: for historical reasons, Sun had a consolidation that was basically only nspr
[16:39] richlowe: shared between the desktop (mozilla), and the JES products.
[16:39] jamesog: Hmm. Now that dev09 is b138 I've noticed some differences with the repo commands
[16:40] richlowe: so you'd need to duplicate that package (see the package of consolidation sic_team)
[16:40] jamesog: there seems to be a pkgrepo command now, and some of the functionality of pkg.depotd has been moved to pkgrepo
[16:40] richlowe: but that's only if you really want to build everything, building media with a mix of your bits and bits from pkg.o.o shouldn't be a problem.
[16:40] Alasdairrr: okie dokie - thanks for the info!

PKG > 141 needed for X

  • No labels