Feature #15665: Cannot compile socket extension on Mojave

• Status: Closed
• Assignee: -
• Target version: -

[franklinyu (Franklin Yu)]

The mkmf.log is reporting that it cannot find netinet6/in6.h. I guess that previously it’s finding it in /usr/include, but as we all know Apple removed that path since Mojave. On my machine the current path is:

/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.14.sdk/usr/include/

The C compiler seems fine if I’m using it stand-alone. I tried compiling following snippet:

#include <netinet/in.h>
int main() {}
int t(struct in6_addr *addr) {
    return IN6_IS_ADDR_UNSPECIFIED(addr);
}

And it works, although I have no idea which header it included. I tried both compiling it with cc test.c, and with the actual command in mkmf.log:

clang -I../../.ext/include/x86_64-darwin18 -I../.././include -I../.././ext/socket -I../.. -I../../. -I/opt/local/include -D_XOPEN_SOURCE -D_DARWIN_C_SOURCE -D_DARWIN_UNLIMITED_SELECT -D_REENTRANT -Wall -Wextra -Wpedantic -Wshadow -pipe -D__APPLE_USE_RFC_3542 -Werror

Both works without any warning/error.

note

This only happens to 2.6.0 and 2.6.1. Releases in 2.5.x or below are not affected.

[hsbt (Hiroshi SHIBATA)]

• Status changed from Open to Rejected

I couldn't reproduce this with Mojave.

~ > ls /usr/include/netinet6/in6.h
/usr/include/netinet6/in6.h

It seems your development environment issue.

[franklinyu (Franklin Yu)]

• Status changed from Rejected to Open

@hsbt (Hiroshi SHIBATA) That location has been deprecated:

The command line tools will search the SDK for system headers by default. However, some software may fail to build correctly against the SDK and require macOS headers to be installed in the base system under /usr/include. If you are the maintainer of such software, we encourage you to update your project to work with the SDK or file a bug report for issues that are preventing you from doing so. As a workaround, an extra package is provided which will install the headers to the base system. In a future release, this package will no longer be provided.

Please be future-proof. It’s totally possible that Ruby 2.6 will break in macOS 10.15.

hsbt (Hiroshi SHIBATA) wrote:

I couldn't reproduce this with Mojave.

~ > ls /usr/include/netinet6/in6.h
/usr/include/netinet6/in6.h

It seems your development environment issue.

[hsbt (Hiroshi SHIBATA)]

• Status changed from Open to Rejected

Do not change status field.

Maybe you can resolve this with sudo installer -pkg /Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg -target /

[nobu (Nobuyoshi Nakada)]

• Status changed from Rejected to Feedback

That check is necessary only for very old, 6 years ago, header.
It is strange that IN6_IS_ADDR_UNSPECIFIED code cannot compile.
Could you show how it failed in mkmf.log?

[franklinyu (Franklin Yu)]

hsbt (Hiroshi SHIBATA) wrote:

Do not change status field.

Maybe you can resolve this with sudo installer -pkg /Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg -target /

That is a workaround. It will break in the future. Please read the bold text in the quote. If by your definition my environment is broken, then maybe next year all macOS user will have broken environment.

nobu (Nobuyoshi Nakada) wrote:

That check is necessary only for very old, 6 years ago, header.
It is strange that IN6_IS_ADDR_UNSPECIFIED code cannot compile.
Could you show how it failed in mkmf.log?

I’m busy now but I will provide the log in 48 hours. Thanks for this information.

[franklinyu (Franklin Yu)]

• File mkmf.log mkmf.log added

nobu (Nobuyoshi Nakada) wrote:

That check is necessary only for very old, 6 years ago, header.
It is strange that IN6_IS_ADDR_UNSPECIFIED code cannot compile.
Could you show how it failed in mkmf.log?

Please find attached. I was trying to install 2.6.3 version.

[nobu (Nobuyoshi Nakada)]

From where -Werror and -Wunused-parameter came?

"clang -I../../.ext/include/x86_64-darwin18 -I../.././include -I../.././ext/socket -I../.. -I../../.  -I/opt/local/include -D_XOPEN_SOURCE -D_DARWIN_C_SOURCE -D_DARWIN_UNLIMITED_SELECT -D_REENTRANT   -Wall -Wextra -Wpedantic -Wshadow -pipe -D__APPLE_USE_RFC_3542  -Werror -c conftest.c"
In file included from conftest.c:1:
In file included from ../.././include/ruby.h:33:
../.././include/ruby/ruby.h:2186:35: error: unused parameter 'allow_transient' [-Werror,-Wunused-parameter]
rb_array_ptr_use_end(VALUE a, int allow_transient)

[nobu (Nobuyoshi Nakada)]

• Tracker changed from Bug to Feature
• ruby -v deleted (2.6.0)
• Backport deleted (2.4: UNKNOWN, 2.5: UNKNOWN, 2.6: UNKNOWN)

[nobu (Nobuyoshi Nakada)]

• Status changed from Feedback to Closed

Applied in changeset git|fe3ff5afb07e171fd950623c69abfbabbb2762a3.

---

Suppress paranoid warnings for external/3rd-party libraries

[Feature #15665]

[franklinyu (Franklin Yu)]

nobu (Nobuyoshi Nakada) wrote:

Applied in changeset git|fe3ff5afb07e171fd950623c69abfbabbb2762a3.

---

Suppress paranoid warnings for external/3rd-party libraries

[Feature #15665]

I have verified that the changeset works for 2.6 on my machine. It is currently in trunk, which would go to 2.7 eventually; would it be backported to 2.6 branch?

[franklinyu (Franklin Yu)]

Hello? Any update? Would it be included in 2.6.4?

[hsbt (Hiroshi SHIBATA)]

I'm not sure nobu's changesets are bug or feature.

But this issue was marked with "Feature" by nobu. We never backport the new feature to stable versions.

[franklinyu (Franklin Yu)]

@nobu (Nobuyoshi Nakada) This should be a bug, not feature. Please mark it as a bug and backport it into 2.6 branch.

[franklinyu (Franklin Yu)]

@nobu (Nobuyoshi Nakada) Please update, or is there any reason that this should be a feature instead of a bug?

[nobu (Nobuyoshi Nakada)]

I did set the tracker to "feature", since this added the support for "-Wpedantic" option, but will leave this to the branch maintainers.

[naruse (Yui NARUSE)]

Usually -pedantic is applied automatically ruby core but not passed to extension libraries.
Such "suppress warnings" are usually not considered as a bug because it may cause side effect.

[franklinyu (Franklin Yu)]

naruse (Yui NARUSE) wrote:

Usually -pedantic is applied automatically ruby core but not passed to extension libraries.
Such "suppress warnings" are usually not considered as a bug because it may cause side effect.

I knew that Ruby was compiled with -pedantic; wasn’t aware that this was for core only. Thanks a lot for the explanation. Look forward to the release in Christmas.

[mame (Yusuke Endoh)]

• Related to Bug #16362: Unable to build Ruby 2.6.5 on Manjaro Linux 18.1.3 added