Opened 16 months ago

Last modified 16 months ago

#67742 assigned defect

PothosFlow: builds fail across-the-board: undefined symbols related to libcxx: allocator<>, basic_string<>, etc

Reported by: mascguy (Christopher Nielsen) Owned by: ra1nb0w
Priority: Normal Milestone:
Component: ports Version: 2.8.1
Keywords: Cc: michaelld (Michael Dickens), mascguy (Christopher Nielsen), barracuda156
Port: PothosFlow

Description

Some notable examples - albeit only a subset - are:

Undefined symbols for architecture x86_64:
  "Poco::Environment::set(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&)", referenced from:
      MyScopedSyslogListener::MyScopedSyslogListener() in PothosFlow.cpp.o

  "Poco::DateTimeFormatter::append(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >&, Poco::DateTime const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int)", referenced from:
      LoggerDisplay::handleLogMessage(Poco::Message const&) in LoggerDisplay.cpp.o

  "Poco::Net::SocketAddress::SocketAddress(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&)", referenced from:
      EnvironmentEval::makeEnvironment() in EnvironmentEval.cpp.o

  "Poco::URI::URI(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&)", referenced from:
      EnvironmentEval::makeEnvironment() in EnvironmentEval.cpp.o

  "Poco::Logger::get(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&)", referenced from:
      myQLogHandler(QtMsgType, QMessageLogContext const&, QString const&) in PothosFlow.cpp.o

[...etc, etc...]

Full log attached, for 10.10 buildbot. Though all appear to be similar.

Attachments (1)

PothosFlow-buildbot-10.10.log.gz (14.7 KB) - added by mascguy (Christopher Nielsen) 16 months ago.

Download all attachments as: .zip

Change History (8)

Changed 16 months ago by mascguy (Christopher Nielsen)

comment:1 Changed 16 months ago by ra1nb0w

thank you for the report. It is related to https://github.com/macports/macports-ports/pull/19424

comment:2 Changed 16 months ago by mascguy (Christopher Nielsen)

Cc: barracuda156 added

Adding Sergey

comment:3 Changed 16 months ago by mascguy (Christopher Nielsen)

Cc: ra1nb0w added; barracuda156 removed
Owner: changed from ra1nb0w to barracuda156

Sergey, can you kindly submit a PR ASAP, with the build fix?

comment:4 Changed 16 months ago by barracuda156

Sorry, just checked Trac. Sure, let me what is going wrong.

comment:5 Changed 16 months ago by barracuda156

UPD. I add correct flags for GCC build here: https://github.com/macports/macports-ports/pull/19636 However, this – as well as https://github.com/macports/macports-ports/pull/19424 – affects only GCC builds. So I cannot see how that could possibly have broken any dependencies on Intel.

I would assume that Macports switched to a newer Clang, or a newer Qt5, or something of the sort, which is incompatible with the old PothosFlow which we have. Judging by commit history with upstream, currently Qt6 is supported. Maybe just update all Pothos* ports to current versions? (PPC does not work at the moment anyway, no need to be concerned about it. I will raise the issue with upstream and see if they suggest something.)

comment:6 in reply to:  3 ; Changed 16 months ago by barracuda156

Replying to mascguy:

Sergey, can you kindly submit a PR ASAP, with the build fix?

So perhaps change the owner to someone who can deal with the matter? I am pretty sure it is not me breaking it, and it is easy to check for someone whose environment has Qt5 and all: drop GCC patch from PothosCore, rebuild, and see if PothosFlow builds. I believe the result should be strictly identical with and without my GCC patch, as long as Clang is used.

comment:7 in reply to:  6 Changed 16 months ago by mascguy (Christopher Nielsen)

Cc: mascguy barracuda156 added; ra1nb0w removed
Owner: changed from barracuda156 to ra1nb0w

Replying to barracuda156:

So perhaps change the owner to someone who can deal with the matter? I am pretty sure it is not me breaking it, and it is easy to check for someone whose environment has Qt5 and all: drop GCC patch from PothosCore, rebuild, and see if PothosFlow builds. I believe the result should be strictly identical with and without my GCC patch, as long as Clang is used.

Sorry Sergey, I was going by comment:1, without digging into the details. Reassigning back to @ra1nb0w.

Note: See TracTickets for help on using tickets.