[blfs-dev] LLVM "compiler-rt" 3.7.0 does not build on 32 bit

Pierre Labastie pierre.labastie at neuf.fr
Tue Sep 8 12:57:13 PDT 2015


On 08/09/2015 16:33, Fernando de Oliveira wrote:
> Em 08-09-2015 10:44, Pierre Labastie escreveu:
>> On 08/09/2015 15:10, Fernando de Oliveira wrote:
>>> Em 08-09-2015 09:18, Pierre Labastie escreveu:
>>>> Hi all,
>>>>
>>>> I've hit the bug reported at
>>>> https://llvm.org/bugs/show_bug.cgi?id=22661. Our friend Armin has found
>>>> a workaround, which I am going to try:
>>>>
>>>> sed -ri "/ifeq.*CompilerTargetArch/s#i386#i686#g"
>>>> projects/compiler-rt/make/platform/clang_linux.mk
>>>>
>>>> I'll test and let you know.
>>>>
>>>> Pierre
>>>>
>>>>
>>> Pierre,
>>>
>>> I didn't update the tests statistics nor tried to see how the docs
>>> install should be changed. Had here a todo for those two issues.
>>>
>>> The tests I had done before I changed something (can't remember anymore)
>>> and ISTR they were fine, so that I didn't wnat to use previous
>>> statistics I had. Tickets were accumulating, and I decided to delay the
>>> fixes.
>>>
>>> Please, would you mind to address those two?
>>>
>>> I know, I'm asking too much things from you, but we are in a hurry, now
>>> and I'm trying to complete the other tickets soon, because some will be
>>> necessary for the tags.
>> OK. Let me have the following ordered todo list:
>> 1) Fix doxygen, sane-backends, 32 bit llvm instructions (within an hour
>> or so). See also the r600 headache in another thread.
>> 2) Update OpenJDK. It is always a lengthy process, because upstream
>> keep changing details, the build/test is 100+SBU, and both a 32 bit and
>> & 64 bit binary have to be built to be hosted on anduin. BTW, I decided
>> to include your download instructions for now. But we should maybe
>> discuss about our policy when upstream packaging is poor: using git
>> or hg or whatever VCS they use can save some headaches. Anyway,
>> this is a discussion for 7.9 (or 8.0 ?).
>> 3) Come back to llvm statistics and docs.
>> 4) Build LFS-7.8-rc1 and do as many tests as possible...
>>
>> Pierre
> 
> Thank you very very much for accepting number 3, besides the others.
> 
> You see I was having a couple of packages with big troubles (there was
> another one that was taking my time, IIRC), and I felt pressed to close
> those tickets and do the accumulating ones, which I'm still doing.
> 

As all the other devs, I see what you accomplish and I am very impressed. I'll
be happy to help for this matter.

Pierre


More information about the blfs-dev mailing list