nghttp2 does not build as +universal

[barracuda156]

So it merely ignores +universal, and that is why it prevented curl from upgrading.

36-61:~ svacchanda$ sudo port -v -n install nghttp2 +universal

--->  Cleaning nghttp2
--->  Removing work directory for nghttp2
36-61:~ svacchanda$ port -v installed nghttp2
The following ports are currently installed:
  nghttp2 @1.47.0_0 (active) requested_variants='' platform='darwin 9' archs='ppc' date='2022-04-24T19:39:31+0800'

[reneeotten (Renee Otten)]

@barracuda156 please do not post large logs of failures as text in a ticket (here and elsewhere), instead add them as attachments

[jmroot (Joshua Root)]

There is no universal variant because there's no compiler that supports ppc, C++14, and universal builds. There would be a message to that effect in the log.

[ryandesign (Ryan Carsten Schmidt)]

So this a duplicate of #48857 essentially.

[barracuda156]

Replying to jmroot:

There is no universal variant because there's no compiler that supports ppc, C++14, and universal builds. There would be a message to that effect in the log.

I guess it has to be done with muniversal PortGroup. There are ports that build as universal using gcc10 and gcc11.

[ryandesign (Ryan Carsten Schmidt)]

Replying to barracuda156:

I guess it has to be done with muniversal PortGroup. There are ports that build as universal using gcc10 and gcc11.

It doesn't have to be done that way, and adding the muniversal portgroup to every port that uses MacPorts gcc on PowerPC systems would probably be a big undertaking, not to mention that using the muniversal portgroup can cause problems of its own. It would be nicer to fix gcc7 and others so that they can build universal "normally", the way that Apple gcc, llvm-gcc and clang do.

[barracuda156]

Replying to ryandesign:

It doesn't have to be done that way, and adding the muniversal portgroup to every port that uses MacPorts gcc on PowerPC systems would probably be a big undertaking, not to mention that using the muniversal portgroup can cause problems of its own. It would be nicer to fix gcc7 and others so that they can build universal "normally", the way that Apple gcc, llvm-gcc and clang do.

I certainly agree with you, but I am not qualified enough to fix gcc. I can only hope that Iain or someone else does this eventually.

[kencu (Ken)]

the gcc driverdriver.c issue has been there for many years now, so nobody seems too motivated to work on it up to now.

Might be more interest now that universal often means arm/intel, so the fix would be more interesting.

[ryandesign (Ryan Carsten Schmidt)]

​https://github.com/macports/macports-ports/pull/24273

[ryandesign (Ryan Carsten Schmidt)]

In 3b3f9b3912f47e8b74a7ca7c6f47032a6f044e6c/macports-ports (master):

nghttp2: Require C++20 only for nghttp2-tools subport

Closes: #65066
Closes: #70027
Closes: #70048