Opened 3 years ago

Closed 3 years ago

Last modified 3 years ago

#61652 closed defect (fixed)

c-ares @1.17.1 does not build on Big Sur X86_64, but the build bot has no problem with arm64 build

Reported by: snowflake (Dave Evans) Owned by: tobypeterson
Priority: Normal Milestone:
Component: ports Version: 2.6.99
Keywords: Cc: A-Petrov-atweb-su, posita (Matt Bogosian)
Port: c-ares

Description

clt version 12.2.0.0.1.1604076827
Xcode 12.2 Build version 12B5025f
Big Sur 11.1 x86_64 20C5048k

The new port configure feature which shows implicit declarations is showing this:

Warning: Configuration logfiles contain indications of -Wimplicit-function-declaration, check that features were not accidentially disabled:
  ioctlsocket: found in config.log
  clone: found in test/config.log
  bitncmp: found in config.log
  close: found in test/config.log
  getuid: found in test/config.log
  usleep: found in test/config.log
  write: found in test/config.log
  CloseSocket: found in config.log
  closesocket: found in config.log

Attachments (3)

c-ares-x86_64_big_sur.log (64.3 KB) - added by snowflake (Dave Evans) 3 years ago.
build log for c-ares x86_64 on Big Sur
c-ares-2020-01-27.main.log.gz (14.4 KB) - added by snowflake (Dave Evans) 3 years ago.
mani.log with verbose logging
console-log-c-ares.txt (1.0 KB) - added by snowflake (Dave Evans) 3 years ago.
console log showing warnings from configure about implicit function declarations

Download all attachments as: .zip

Change History (14)

Changed 3 years ago by snowflake (Dave Evans)

Attachment: c-ares-x86_64_big_sur.log added

build log for c-ares x86_64 on Big Sur

comment:1 Changed 3 years ago by jmroot (Joshua Root)

Needs to disable silent build rules, but this could be the usual libtool bug of not recognising 11.x as a macOS version.

comment:2 Changed 3 years ago by A-Petrov-atweb-su

Cc: A-Petrov-atweb-su added

comment:3 Changed 3 years ago by posita (Matt Bogosian)

Cc: posita added

comment:4 Changed 3 years ago by tobypeterson

Sorry for the delayed response. I'm not able to reproduce this failure.

comment:5 Changed 3 years ago by tobypeterson

As suggested by jmr above, I've pushed a change to disable silent rules (i.e. enable verbose build output). If this still reproduces, can you provide logging with that change?

comment:6 Changed 3 years ago by snowflake (Dave Evans)

As the original poster, I find that c-ares has been installed here on Big Sur sometime within the last two months, probably by my script which installs ports listed on my El Capitan Mac. It is up to date and was probably installed from an archive.

I still can't build it, but there is probably something funny about my OS. I notice that the Macports buildbot is building it OK as of 2020-12-08 16:13:43.

I suggest this ticket be closed and we'll put it down to a curiosity of my system. I will attach logs.

Changed 3 years ago by snowflake (Dave Evans)

mani.log with verbose logging

Changed 3 years ago by snowflake (Dave Evans)

Attachment: console-log-c-ares.txt added

console log showing warnings from configure about implicit function declarations

comment:7 Changed 3 years ago by snowflake (Dave Evans)

I'll point out that my system is macOS 11.2 (darwin/20.3.0) arch i386 which is probably not the same as the buildbot's

comment:8 Changed 3 years ago by tobypeterson

Looks like this port fails to build if gtest is installed.

c-ares includes its own copy of gtest, but it looks like we end up getting the headers from the gtest port instead, which have macros referencing symbols that don't appear in c-ares' copy of gest.

comment:9 Changed 3 years ago by tobypeterson

Resolution: fixed
Status: assignedclosed

In e293c499521af020b40b550c801f0d0aa23d9844/macports-ports (master):

c-ares: avoid MacPorts headers, fixes #61652

comment:10 Changed 3 years ago by ryandesign (Ryan Carsten Schmidt)

Version: 2.5.992.6.99

But that hasn't addressed the implicit function declarations in the configure tests that may be causing them to return wrong results.

comment:11 Changed 3 years ago by tobypeterson

Yeah, but it's just in the tests. Besides, there's now a PR moving to cmake anyway, so that'll be even less relevant once that goes in.

Note: See TracTickets for help on using tickets.