Re: [Ltib] Issue when changing toolchains in a build

[Peter Barada]

On 10/04/2012 08:13 AM, Peter Barada wrote:
> It's other packages...

To be specific, its _any_ prepped package that was built using the older
toolchain.  Yes, I see how the kernel handles this case.  I *could* add
into every .spec file's %build stanza a test for ${LTIB_FULL_REBUILD}
and if set then do "make distclean", but since LTIB already has code in
place to remove all the packages I figured it would be easier during the
RPM removal to see if the builddir existed and if so execute "make
distclean" to clean it out (ignoring the result of make in cases where
they package doesn't have a proper distclean make stanza).
> Splat-fingered from my iPhone
>
> On Oct 4, 2012, at 2:29 AM, "Stuart Hughes" <address@hidden> wrote:
>
>> On 03/10/12 17:45, Peter Barada wrote:
>>> On 10/03/2012 12:08 PM, Stuart Hughes wrote:
>>>> On 03/10/12 16:46, Peter Barada wrote:
>>>>> Stuart,
>>>>>
>>>>> If I have a build where the kernel source is needed to build another
>>>>> module, (or some other package source is needed for another package to
>>>>> build), then the package is built and the source/object files are in
>>>>> rpm/BUILD.  If I then do "./ltib -c" and change the toolchain, LTIB
>>>>> properly de-installs all the packages, and then builds the packages
>>>>> again using the new toolchain, including the source that was left behind
>>>>> from the previous build.
>>>>>
>>>>> So far so good, but if the new toolchain default (on ARM for example)
>>>>> causes incompatible object files (i.e. using hard vs soft float ABI on
>>>>> ARM), then the package source that was previously built will fail to
>>>>> build/link due to the incompatibility.
>>>>>
>>>>> I think the right way to fix this is to execute a "distclean" on package
>>>>> source that exist, triggered by ${LTIB_FULL_REBUILD} before rebuilding,
>>>>> but so far I haven't quite figured out the path to get this to happen.
>>>>> I'm more than happy to tackle it but need a pointer that makes the code
>>>>> flow clear.
>>>>>
>>>>> Thanks in advance!
>>>>>
>>>> Hi Peter,
>>>>
>>>> I agree with your analysis.  Again I need to go off and think about
>>>> this dark corner of LTIB.  Once I've refreshed my memory, I'll get
>>>> back to you.
>>> Thanks!
>>>
>> Hi Peter,
>>
>> In the case of kernels, they should already do a full clean if a 
>> LTIB_FULL_REBUILD is set.  If you look in 
>> ./dist/lfs-5.1/kernel/kernel-common.tmpl line 114, you'll see:
>>
>> #
>> # Check for ltib full rebuilds (e.g. change of toolchain) if so
>> # force a build from scratch
>> #
>>
>> if [ -n "$LTIB_FULL_REBUILD" ]
>> then
>>     make ARCH=$LINTARCH HOSTCC="$BUILDCC" mrproper
>> fi
>>
>> Is this causing a problem/not working, or is it other packages that you 
>> need to distclean?
>>
>> Regards, Stuart
>>
>>
>>
>>
>>
>> _______________________________________________
>> LTIB home page: http://ltib.org
>>
>> Ltib mailing list
>> address@hidden
>> https://lists.nongnu.org/mailman/listinfo/ltib
> _______________________________________________
> LTIB home page: http://ltib.org
>
> Ltib mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/ltib


-- 
Peter Barada
address@hidden