[OE-core] [PATCH 2/6] gmp: Bring back version 4.2.1 (LGPL 2.1+)

[Mark Hatle]

On 8/31/15 10:27 AM, Khem Raj wrote:
> 
>> On Aug 31, 2015, at 8:21 AM, Martin Jansa <martin.jansa at gmail.com> wrote:
>>
>> On Mon, Aug 31, 2015 at 10:22:31AM -0300, Otavio Salvador wrote:
>>> On Sat, Aug 29, 2015 at 7:18 AM, Richard Purdie
>>> <richard.purdie at linuxfoundation.org> wrote:
>>>> On Sat, 2015-08-29 at 10:43 +0200, Martin Jansa wrote:
>>>>> Did you enable thumb in that qemuarm build?
>>>>>
>>>>>
>>>>> It's disabled by default and multiple components are broken when you
>>>>> enable it, see:
>>>>> https://bugzilla.yoctoproject.org/show_bug.cgi?id=7717
>>>>
>>>> I'm going to draw a line here. People are asking that the old version of
>>>> gmp be brought back due license issues, fine. It doesn't mean that all
>>>> the bugs in said old release should be fixed.
>>>>
>>>> If thumb on arm with old gmp is an issue for people, I'd suggest that
>>>> those people figure out which patches are needed and send the patches.
>>>> Asking Jussi to deal with this isn't right/fair though.
>>>
>>> I disagree. If it will be merged it ought to support what we are
>>> testing nowadays otherwise the backlog will be kept growing forever...
>>
>> FWIW: I don't mind getting it merged as is.
>>
>> It improves situation for non-GPLv3 builds and if there is issue with
>> thumb, then it was there before as well, so returning it doesn't make
>> things worse.
>>
>> My previous reply was just to show that building default qemuarm doesn't
>> show anything, because thumb isn't enabled and gmp even isn't one of the
>> recipes which are failing with thumb-1, maybe Khem meant some issues in
>> runtime I don't know about.
> 
> It has issues working with thumb2 as well. Another problem it creates is building on-device toolchain
> since gcc also depends on gmp and when you pin to older version of GMP, it starts to fail
> Thats why I suggested to use mini-gmp ( same but reduced gmp ) can clear of this situation since it
> will be a mini-gmp recipe and won’t interfere with toolchain.
> 

Do you have any sense of the problems when gcc uses an older version of GMP?  If
a newer version is required for gcc -- then can gcc itself check that it's a
proper minimum version available?  (At this point I would assume if someone is
worried about GPLv3 issues, that they would NOT be putting the compiler and
related tools on the target.)  I know we can specify -runtime- minimum versions
easily enough using the > directive.. but I don't think there is an easy way to
specify minimum build version, other then verify we have the correct version
available to us though a file check of some kind.

--Mark