Cannot build libb2

[Ryan Schmidt]

On Dec 15, 2021, at 14:25, Dave Horsfall wrote:

> On Fri, 10 Dec 2021, Ryan Schmidt wrote:
> 
>> The "-arch x86_64 -arch i386" in the log does seem to indicate MacPorts 
>> is doing a universal build. The ways that could happen:
>> 
>> * You specified +universal on the command line
> 
> Nope...
> 
>> * You specified +universal in variants.conf
> 
> Nope...
> 
>> * You installed a port that, because it does not support your machine's 
>> native architecture, requires its dependencies to be installed 
>> universal.
> 
> Could be; how do I tell which variants were used by what?

If you are installing a port that indicates, for example, that it only supports i386, and it depends on other ports, then MacPorts must install those other ports universal for i386 and x86_64. If it did not, then you would only have an x86_64 version of them, which wouldn't be sufficient for whichever port it is that only supports i386.

To discover which of your installed ports fit that description, you could run:

port -qv installed | grep -Ev 'archs=.*noarch|x86_64'


>> If libb2 does not build universal, then that is certainly something we 
>> should fix. A bug has already been filed about it:
>> 
>> https://trac.macports.org/ticket/64148
> 
> Thanks; I'll try the suggested fix, but...
> 
>> If you want to explore why libb2 is building universal when you don't 
>> intend that, show us what command you ran and the output. Specifically, 
>> we'd want to know which port you are trying to install or upgrade when 
>> libb2 gets build universal.
> 
> It was no more than my weekly "port selfupdate; port upgrade outdated" 
> etc.

Great, so one of your outdated ports, or one of its dependencies, is causing libb2 to build universal. We'd need to see the output of your commands, possibly with debug output, to know which one, if the above `port -qv installed ...` command doesn't provide the answer.


> Looks like I forgot to run it under "script", but when I tried again it 
> now mysteriously works, so I dunno...  Chalk it up to a "heisenbug", I 
> guess (I hate those).

If you selfupdated or synced before trying again, then your success is expected, since I fixed the problem as indicated in #64148.