Re: ANNOUNCE: textproc/icu updates are now restricted after 2/15, in quarterly

[Ryo ONODERA <ryo%tetera.org@localhost>]

Hi,

Greg Troxel <gdt%lexort.com@localhost> writes:

> Ryo ONODERA <ryo%tetera.org@localhost> writes:
>
>> Hi,
>>
>> Greg Troxel <gdt%lexort.com@localhost> writes:
>>
>>> Ryo ONODERA <ryo%tetera.org@localhost> writes:
>>>
>>>> Some dependencies of misc/libreoffice including fonts/harfbuzz-icu require
>>>> the similar things.
>>>> misc/libreoffice itself requires relatively big patch for ICU 75 and I will
>>>> commit the fixes tonight hopefully.
>>>>
>>>> As far as I remember correctly, x11/qt5-qtbase has alloca() problem when
>>>> -std=c++NN is specified.
>>>> ICU itself may be safe for -std=c++NN, however it may be not safe totally.
>>>
>>> I realize you aren't the one that updated icu, but this seems like quite
>>> a lot of problems.
>>>
>>> If we can't get things fixed soon, say in the next 7 days, I think need
>>
>> Could you please describe what problems should be resolved?
>>
>> misc/libreoffice should be built without problem.
>> Of course the other packages has similar problems and they should be
>> fixed.
>
> It seems that this was an API break that requires accomodations in a
> number of packages, and you've just committed changes to libreoffice.
> That's great that you fixed this.
>
> What I meant is that there is a general expectation that when an update
> is committed, everything that worked before should continue to build and
> work.  That means that before this kind of API break lands, the
> depending pacakges should be upgraded before, not after, and they of
> course should be able to build with either icu.  Their upstreams should
> have already issued a release, because icu should have done proper
> notice about these API breaks.
>
> I don't know how many other packages are troubled and need adjusting,
> but it seems reasonably likely to be "everything that depends on icu".
> So we will have to see how much breakage remains.

Thanks for your explanation.

I agree. The cares for ABI breakage should be prepared before update.
I will do my best to fix build breaks caused by ICU 75.

And we should have meta-pkgs/bulk-test-icu.
I will create it.

>> The problem between -std=c++NN and -std=gnu++NN is long standing
>> problem and it is not related to ICU 75 at all.
>
> OK, but that seems not what is causing the real trouble.

Thank you.

-- 
Ryo ONODERA // ryo%tetera.org@localhost
PGP fingerprint = 82A2 DC91 76E0 A10A 8ABB  FD1B F404 27FA C7D1 15F3