AnRe: latest update to BigSur 11.1 broke gcc

[Chris Jones]

I tried it myself, and switching to --with-build-sysroot with gcc10 on big sur does seem to build. I haven't had time to run any tests to see exactly what this means w.r.t. the default sdk search paths. But in any case I think the best advice is anyway to explicitly set it yourself, using one of the numerous methods (command line flag, SDKROOT, or use xcrun) in which case whatever the default search paths  are does not really make much difference. If we do decide to switching to the above then what this means in practise I think is just it enforces that users will have to do this, rather than it sometimes working for them.

Out of interest what happens with the macports clang ports ? If no default path to an sdk is set there, presumably users are also required with these compilers to always explicitly give the sdk they wish to use, via similar methods ?

Chris 

> On 16 Dec 2020, at 3:42 pm, Ken Cunningham <ken.cunningham.webuse at gmail.com> wrote:
> 
> 
> I'll try it. Things have changed in gcc -- after all, we don't bake any such path into our clang installs, and they don't exhibit this issue.
> 
> K
> 
>> On Dec 16, 2020, at 02:38, Christopher Jones <jonesc at hep.phy.cam.ac.uk> wrote:
>> 
>> From a combination of my memory, and reading the tickets referenced in the gcc port file, if we don’t use 
>> —with-sysroot we would need to use --with-build-sysroot in order for the build to work, and last time that was tried it didn’t work correctly.
>> 
>>> On 16 Dec 2020, at 3:37 am, Ken Cunningham <ken.cunningham.webuse at gmail.com> wrote:
>>> 
>>> we might just delete this line from the portfile, perhaps:
>>> 
>>> 
>>>  configure.args-append --with-sysroot="${configure.sdkroot}"
>> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.macports.org/pipermail/macports-users/attachments/20201216/22c2a935/attachment.htm>