I'm starting to prepare for the armv7hl koji builders... a few thoughts:
- I don't think we can have builders that do both armv5tel and armv7hl, because the mock environment is populated with packages from outside the chroot (i.e., mock populates the directory and then does the chroot). Thus, any scriptlets will execute in the wrong environment. This should be a rare occurrence, but since hardfp and softfp happily link together and *do the wrong thing*, we need to avoid this scenario.
- We haven't planned to do multiarch for arm, but it might be desired in some contexts, particularly if proprietary software is available only with a particular ABI (e.g., Flash in softfp only?). Multiarch might also permit builders to do both armv5tel and armv7hl. I think the cost (effort, storage space) probably outweighs the benefits but the discussion is worthwhile.
- It would be really nice to get ld patched so that attempting to link hardfp and softfp would produce an error, instead of silently producing undefined results.
- I'd like to build F15 on both armv5tel and armv7hl at the same time, i.e., submitting to a.k.fo.o with a target of dist-f15 will kick off a build on both variants by default (like 32-/64-bit builds on other archs). To this end, Paul Whalen is looking at putting together an armv5tel seed package set that matches the emerging armv7hl seed package set as closely as possible.
-- Chris
On Thu, Jul 7, 2011 at 3:08 PM, Chris Tyler chris@tylers.info wrote:
I'm starting to prepare for the armv7hl koji builders... a few thoughts:
- I don't think we can have builders that do both armv5tel and armv7hl,
because the mock environment is populated with packages from outside the chroot (i.e., mock populates the directory and then does the chroot). Thus, any scriptlets will execute in the wrong environment. This should be a rare occurrence, but since hardfp and softfp happily link together and *do the wrong thing*, we need to avoid this scenario.
- We haven't planned to do multiarch for arm, but it might be desired in
some contexts, particularly if proprietary software is available only with a particular ABI (e.g., Flash in softfp only?). Multiarch might also permit builders to do both armv5tel and armv7hl. I think the cost (effort, storage space) probably outweighs the benefits but the discussion is worthwhile.
- It would be really nice to get ld patched so that attempting to link
hardfp and softfp would produce an error, instead of silently producing undefined results.
- I'd like to build F15 on both armv5tel and armv7hl at the same time,
i.e., submitting to a.k.fo.o with a target of dist-f15 will kick off a build on both variants by default (like 32-/64-bit builds on other archs). To this end, Paul Whalen is looking at putting together an armv5tel seed package set that matches the emerging armv7hl seed package set as closely as possible.
I've already done some work on the armv5tel F-15 packaging but as we should rebuild all the hardfp packages within koji anyway from the initial out of koji builds I think we should be able to do them both together within koji without needing to do it separately.
Peter
I have been watching the build process closely so that I could try to do a arm5 build. I have a sheeva plug and I have a f12 and f13 build I have been working with to try to bootstrap my own f15 from looking at what they are doing . If you have some instructs and a already compiled stage 1 I can chroot into I would be more than happy to participate in building rpms from srpms and creating a rootfs. I'm new at this and watching each week closely to see how this all works :) give me info and I'll run with it as much as I can!!!
Sent from my iPhone
On Jul 7, 2011, at 10:16 AM, Peter Robinson pbrobinson@gmail.com wrote:
On Thu, Jul 7, 2011 at 3:08 PM, Chris Tyler chris@tylers.info wrote: I'm starting to prepare for the armv7hl koji builders... a few thoughts:
- I don't think we can have builders that do both armv5tel and armv7hl,
because the mock environment is populated with packages from outside the chroot (i.e., mock populates the directory and then does the chroot). Thus, any scriptlets will execute in the wrong environment. This should be a rare occurrence, but since hardfp and softfp happily link together and *do the wrong thing*, we need to avoid this scenario.
- We haven't planned to do multiarch for arm, but it might be desired in
some contexts, particularly if proprietary software is available only with a particular ABI (e.g., Flash in softfp only?). Multiarch might also permit builders to do both armv5tel and armv7hl. I think the cost (effort, storage space) probably outweighs the benefits but the discussion is worthwhile.
- It would be really nice to get ld patched so that attempting to link
hardfp and softfp would produce an error, instead of silently producing undefined results.
- I'd like to build F15 on both armv5tel and armv7hl at the same time,
i.e., submitting to a.k.fo.o with a target of dist-f15 will kick off a build on both variants by default (like 32-/64-bit builds on other archs). To this end, Paul Whalen is looking at putting together an armv5tel seed package set that matches the emerging armv7hl seed package set as closely as possible.
I've already done some work on the armv5tel F-15 packaging but as we should rebuild all the hardfp packages within koji anyway from the initial out of koji builds I think we should be able to do them both together within koji without needing to do it separately.
Peter
arm mailing list arm@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/arm
-
Chris Tyler -
Donald Gullett -
Peter Robinson