On 19-09-15 18:04, Gordan Bobic wrote:
> On 19/09/15 16:25, Jacco Ligthart wrote:
>>>> This looks like a nice description of what I see, including a patch
>>>> for glibc:
>>>> http://permalink.gmane.org/gmane.linux.linaro.cross-distro/121
>>>>
>>>> Any thoughts on this?
>>> To respond to myself:
>>> I got TB to build. the only changes made were Gordan's gcc-4.8 patch
>>> and the above mentioned glibc patch.
>>> If the mirrors pick it up, you guys can test. (I'm away from a test
>>> machine. I can build, but no clue if it installs, let alone works)
>>>
>>> FF is building right now. we have to wait to see if the same is
>>> enough to build that one also.
>>
>> And it failed again, but differently. I'm back to the errors I reported
>> earlier:
>>
>> <error>
>> In file included from
>> /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/objdir/js/src/Unified_cpp_js_src6.cpp:38:0:
>>
>> /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/Architecture-arm.cpp:
>>
>> In function 'uint32_t js::jit::ParseARMCpuFeatures(const char*, bool)':
>> /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/Architecture-arm.cpp:80:22:
>>
>> error: 'HWCAP_VFPv4' was not declared in this scope
>>               flags |= HWCAP_VFPv4;
>>                        ^
>> /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/Architecture-arm.cpp:84:22:
>>
>> error: 'HWCAP_IDIVT' was not declared in this scope
>>               flags |= HWCAP_IDIVT;
>> </error>
>>
>> I thought that these errors were similar to Gordan's. So I tried all the
>> patches and config changes which worked for FF 24 (of course they did
>> not apply cleanly). End result after many tries and retries, also with
>> all these patches I'm back to the error message of above :(
>
> Really? Despite the patch from back in FF 24 that makes it disable VFP
> stuff on ARM (since we're soft-float)? That's... quite annoying...
>
>> I'm not sure how to continue this the smart way. First, however, I'm
>> going to try the dumb way: build the rpm on a rpi2 (vs the odroid I
>> usually use for building) the rpi2 *does* have the vfpv4 and idivt flags
>> in its features.
>
> I suspect the problem is more fundamental than that. Seems there were
> assumptions made about ARM always being hard-float, which is what the
> old patches in our FF 24 package were supposed to address.
indeed that also did not work. I still wonder what's the difference
between FF on RSEL6 and 7. why does it work normally on 7?

>
> Is there a way to disable JIT alltogether? That might bypass the whole
> issue.
Not sure, when I google for that I get the idea that there is a flag to
disable, but also that it does not work as advertised :(

Jacco
_______________________________________________
users mailing list
[email protected]
http://lists.redsleeve.org/mailman/listinfo/users

Reply via email to