py311-scipy now builds for me with Xcode 15.1 beta. So the solution IMHO will 
be to wait for the Xcode 15.1 release and unbreak gcc13 to use ld instead of 
ld_classic. 


-Frank

> On Oct 21, 2023, at 6:47 AM, Frank Schima <[email protected]> wrote:
> 
> I was finally able to build boost176 and now I’m stuck trying to build 
> py311-scipy. I’m following #68329. 
> 
> 
> -Frank
> 
>> On Oct 20, 2023, at 11:02 AM, Christopher Jones <[email protected]> 
>> wrote:
>> 
>> 
>> boost176 not building is a different issue to the one I was talking about, 
>> which was the linker issue affecting a number of builds. So no, I don’t 
>> necessarily expect Xcode 15.1 to help there.
>> 
>> Is there a ticket open with your problem with boost ? If not please open one.
>> 
>> Chris
>> 
>>> On 20 Oct 2023, at 5:04 pm, Frank Schima <[email protected]> wrote:
>>> 
>>> Correct. I installed the beta CLT (as I mentioned). 
>>> 
>>> No joy. boost176 does not build. 
>>> 
>>> 
>>>> On Oct 20, 2023, at 2:26 AM, Christopher Jones <[email protected]> 
>>>> wrote:
>>>> 
>>>> 
>>>> Do you also have CLT installed and updated to 15.1 as well (if there is a 
>>>> version) ?
>>>> 
>>>> If you still have 15.0 CLT then likely it won’t help.
>>>> 
>>>>> On 20 Oct 2023, at 1:23 am, Frank Schima <[email protected]> wrote:
>>>>> 
>>>>> py311-scipy still does not build for me with the latest Xcode 15.1 beta 
>>>>> and associated CLT. I can only hope I’m doing something wrong. 
>>>>> 
>>>>> I cannot even get past boost176 which is required. I modified my local 
>>>>> Portfile to use boost181 (which does build) but py311-scipy still does 
>>>>> not build. 
>>>>> 
>>>>> 
>>>>> 
>>>>>> On Oct 19, 2023, at 1:11 PM, Christopher Jones 
>>>>>> <[email protected]> wrote:
>>>>>> 
>>>>>> The fix is in Xcode 15.1, not 15.0.1
>>>>>> 
>>>>>> Chris
>>>>>> 
>>>>>>> On 19 Oct 2023, at 5:27 pm, Jim Secan <[email protected]> wrote:
>>>>>>> 
>>>>>>> I see that there’s an Xcode 15.0.1 released yesterday.  Anyone tried 
>>>>>>> that?
>>>>>>> 
>>>>>>> Jim
>>>>>>> Seattle, WA
>>>>>>> 
>>>>>>>> On 10/19/2023, at 8:00 AM, Chris Jones <[email protected]> 
>>>>>>>> wrote:
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>> On 19/10/2023 3:59 pm, Chris Jones wrote:
>>>>>>>>> On 19/10/2023 3:09 pm, Artemio González López via macports-users 
>>>>>>>>> wrote:
>>>>>>>>>> I just realized that my problem could exactly be the one reported in 
>>>>>>>>>> ticket #68329 (py311-scipy @1.10.1_0+gfortran+openblas not building 
>>>>>>>>>> on Sonoma apple silicon). Indeed, I just spotted the following lines 
>>>>>>>>>> in my main.log:
>>>>>>>>>> 
>>>>>>>>>> :info:build ld: duplicate LC_RPATH '/opt/local/lib/libgcc' in 
>>>>>>>>>> '/opt/local/lib/libopenblas-r1.dylib'
>>>>>>>>>> :info:build clang: error: linker command failed with exit code 1 
>>>>>>>>>> (use -v to see invocation)
>>>>>>>>>> 
>>>>>>>>>> Short of waiting for Apple to release XCode 15.1, is there any quick 
>>>>>>>>>> fix for this?
>>>>>>>>> The canocial fix is to tell the build to use the classic linker 
>>>>>>>>> option. 'Well behaved' builds will respect the following setting in 
>>>>>>>>> the Portfile
>>>>>>>>> configure.ldflags-append  -Wl,-ld_classic
>>>>>>>>> This has worked for me in a few places, like the root6 port. I did 
>>>>>>>>> actually try it in py-scipy and there it did not work, and nor did a 
>>>>>>>>> handful of other tricks I have to get the build to use a certain 
>>>>>>>>> linker.
>>>>>>>>> In my experience python based builds are not that well behaved when 
>>>>>>>>> it comes to things like respecting the macports build flags...
>>>>>>>>> At this point I am in the 'wait for Xcode 15.1' camp. Usually the 
>>>>>>>>> first X.1 OS release comes along relatively soon after the initial 
>>>>>>>>> X.0 release, so I would anticipate it relatively soon, most likely.
>>>>>>>> 
>>>>>>>> b.t.w. If it is urgent for you, there is I believe a beta version of 
>>>>>>>> Xcode 15.1 available you could try. Its not the urgent for me so I am 
>>>>>>>> just waiting for it to become public..
>>>>>>>> 
>>>>>>>>> cheers Chris
>>>>>>> 
>>>>>> 
>>>>> 
>>>> 
>>> 
>> 
> 

Reply via email to