[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Can't create an i386 ISO, but amd64 works fine



On 11-07-14 09:28 PM, Ben Armstrong wrote:
On 14/07/11 06:39 PM, Daniel Baumann wrote:
On 07/14/2011 07:06 PM, Ben Armstrong wrote:
if you do both arches from the same tree there would be a risk that one image is
"tainted" with stuff from the wrong arch if you didn't take care.

provided that one does a) ensure the generated config files in config/
are removed in auto/clean, b) the generated config files in config/* are
created in auto/config, and c) one does a lb clean before 'switching'
the architecture, i see no problem in having two archs configured in one
tree.

Not just lb clean, but lb clean --purge to get rid of cached bootstrap.
That's why I'm not really keen on this approach, as if you're doing much
switching between architectures, you'll lose the benefit of keeping the
cache around. Now, if the caches existed in arch-specific directories,
that would be another matter ... I haven't checked this ...

This is how I assumed it worked in the first place, especially with the option being called "LB_ARCHITECTURES". I thought I would be able to specify multiple architectures like this:

    --architectures i386 amd64

and have them built by the one invocation (build platform-permitting).

just to be clear: while i keep as described all architectures and
distributions in one single config tree, i do check it out on the buildd
freshly from git every time i want to build it. so, tainting is not so
much of a problem.

Ah, so you're already not benefiting from caching.

otoh, if the config tree is not done well, the result can be that either
the different architectures images are not really aequivalent (same
package selection, etc), or, that the image can actually be built. so,
as long as one makes sure that all the arch conditionals are configured
properly for all architectures aequivalently, the images if built in a
proper environment, will be fine.

Sure. Well, I don't fundamentally disagree with you on any point of your
approach, but as things stand I don't think there is one "ideal"
solution. I guess it just depends on what tradeoffs you prefer.

Having separate caches kept for each architecture would mitigate the caching issue. Perhaps that could be considered for a future release? This would allow people to have one "code base" for all architectures, simplifying the process and avoiding mistakes when trying to synchronize between architectures.


Reply to: