Here
<https://indiestack.com/2023/10/xcode-15-duplicate-library-linker-warnings/>
is a blog post which appears to corroborate my hypothesis that the
duplicate libraries are being generated by XCode 15 and not by Sage's build
system.
If the warnings are causing test failures it would probably be best to
suppress them for now and somehow remember to re-enable them once the XCode
bug gets fixed.
I still do not understand why adding -Wl,-ld_classic to LDFLAGS in sage-env
does not cause the old linker to be used when building some spkgs.
Apparently the LDFLAGS environment variable is not used in those cases, but
I don't know why not.
- Marc
On Thursday, April 25, 2024 at 11:30:52 PM UTC-5 Marc Culler wrote:
> This looks pretty complicated.
>
> I used XCode 15.0.1 to build the macOS SageMath-10-3 app for Arm and XCode
> 13.4 for Intel. I got "duplicate rpaths" and "duplicate libraries"
> warnings with 15.0.1 but not with 13.4. I pasted typical gcc commands (from
> cvxopt) below. You can see that:
> (1) Both commands contain duplicate rpath specifications but XCode 13.4
> does not warn about them
> (2) duplicate libraries and invalid library search paths only occur with
> XCode 15.01.
> (3) With 15.0.1 the gcc command does not include the linker option
> -ld_classic, so the code in sage-env is not actually succeeding in using
> the old linker. (But the build works anyway.)
>
> It is not clear to me that Sage is responsible for the duplicates, since
> it seems to depend on the XCode version, and it is not clear to me that the
> code in sage-env that aims to use the old linker has any effect, on spkgs
> anyway.
>
> - Marc
>
> XCode 15.0.1 on Arm:
> -----------------------------------
> gcc -bundle -undefined dynamic_lookup
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-platform_version,macos,11.0,11.1
> -L/Library/Developer/CommandLineTools/SDKs/macOSX.sdk/usr/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/var/lib/sage/venv-python3.11.1/lib
>
> -L. -Wl,-rpath,. -L/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-platform_version,macos,11.0,11.1
> -L/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-platform_version,macos,11.0,11.1
> -L/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib -O2
> -mmacosx-version-min=11.0
> build/temp.macosx-11.0-arm64-cpython-311/src/C/base.o
> build/temp.macosx-11.0-arm64-cpython-311/src/C/dense.o
> build/temp.macosx-11.0-arm64-cpython-311/src/C/sparse.o
> -L/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/var/lib/sage/venv-python3.11.1/lib
> -lm -lopenblas -lopenblas -o build/lib.macosx-11.0-arm64-cpython-311/cvxopt/
> base.cpython-311-darwin.so
> ld: warning: duplicate -rpath
> '/private/var/tmp/sage-10.3-current/local/lib' ignored
> ld: warning: duplicate -rpath
> '/private/var/tmp/sage-10.3-current/local/lib' ignored
> ld: warning: duplicate -rpath
> '/private/var/tmp/sage-10.3-current/local/lib' ignored
> ld: warning: duplicate -rpath
> '/private/var/tmp/sage-10.3-current/local/lib' ignored
> ld: warning: duplicate -rpath
> '/private/var/tmp/sage-10.3-current/local/lib' ignored
> ld: warning: ignoring duplicate libraries: '-lopenblas'
> ld: warning: search path
> '/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib' not found
> ld: warning: search path
> '/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib' not found
> ld: warning: search path
> '/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib' not found
> ld: warning: duplicate LC_RPATH are deprecated
> ('/private/var/tmp/sage-10.3-current/local/lib')
>
> XCode 13.4 on Intel:
> --------------------------------
> gcc -bundle -undefined dynamic_lookup
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-platform_version,macos,10.9,11.3
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/var/lib/sage/venv-python3.11.1/lib
>
> -L. -Wl,-rpath,. -L/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-platform_version,macos,10.9,11.3
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-rpath,/private/var/tmp/sage-10.3-current/local/lib
> -Wl,-platform_version,macos,10.9,11.3 -O2 -mmacosx-version-min=10.9
> -mno-avx -mno-avx2 -mno-bmi2
> build/temp.macosx-10.9-x86_64-cpython-311/src/C/base.o
> build/temp.macosx-10.9-x86_64-cpython-311/src/C/dense.o
> build/temp.macosx-10.9-x86_64-cpython-311/src/C/sparse.o
> -L/private/var/tmp/sage-10.3-current/local/lib
> -L/private/var/tmp/sage-10.3-current/local/var/lib/sage/venv-python3.11.1/lib
> -lm -lopenblas -lopenblas -o
> build/lib.macosx-10.9-x86_64-cpython-311/cvxopt/base.cpython-311-darwin.so
> ld: warning: passed two min versions (10.9, 10.9) for platform macOS.
> Using 10.9.
> ld: warning: passed two min versions (10.9, 10.9) for platform macOS.
> Using 10.9.
> ld: warning: passed two min versions (10.9, 10.9) for platform macOS.
> Using 10.9.
>
> - Marc
>
>
> On Thursday, April 25, 2024 at 3:19:50 PM UTC-5 John H Palmieri wrote:
>
>> Take a look at the discussion at
>> https://github.com/sagemath/sage/pull/36337, in particular
>> https://github.com/sagemath/sage/pull/36337#issuecomment-1741293729. I
>> saw both "duplicate rpath" warnings and "duplicate library" warnings. Some
>> (maybe all, at least at that point) of the duplicate rpaths were removed by
>> https://github.com/sagemath/sage/pull/36364, but I don't know how to
>> prevent the duplicate libary warnings.
>>
>> --
>> John
>>
>>
>> On Thursday, April 25, 2024 at 1:07:17 PM UTC-7 marc....@gmail.com wrote:
>>
>>> Hi John,
>>>
>>> I have noticed that .dylib files generated by Sage often have as many as
>>> 3 identical rpaths. (When building the macOS app I remove all rpaths and
>>> replace them with a single rpath which is relative, meaning it starts with
>>> @loader_path. Apple will not notarize app bundles with absolute rpaths.)
>>>
>>> I wonder if the "duplicate libraries" warnings are caused by duplicate
>>> rpath entries in the mach files. I will try to test that hypothesis.
>>>
>>> Do you have any idea what could cause those duplicate rpaths? Are there
>>> multiple -rpath options being passed to the compiler?
>>>
>>> - Marc
>>>
>>> On Thu, Apr 25, 2024 at 2:50 PM John H Palmieri <jhpalm...@gmail.com>
>>> wrote:
>>>
>>>> Hi Marc,
>>>>
>>>> I just tried building Sage without `-ld-classic`. It builds, but I get
>>>> warnings about "ignoring duplicate libraries", and those cause doctest
>>>> failures. The lines could be modified to test whether xcode-select is
>>>> present and executable first, or since Sage now does indeed build without
>>>> `-ld-classic`, we could filter out the warnings when doctesting.
>>>>
>>>> --
>>>> John
>>>>
>>>>
>>>> On Wednesday, April 24, 2024 at 3:20:54 PM UTC-7 marc....@gmail.com
>>>> wrote:
>>>>
>>>>> Well, it almost solved the problem.
>>>>>
>>>>> It turns out that calling /usr/bin/gcc was not the only issue in
>>>>> sage-env. That script also calls xcode-select. On a system with no
>>>>> XCode
>>>>> app and no command line tools, calling gcc causes an error message to be
>>>>> printed to stderr and a dialog to open asking whether to install the
>>>>> command line tools. Calling xcode-select on such a system prints the
>>>>> same
>>>>> error message but does not open the dialog. The error message appears in
>>>>> the terminal when running sage in a command line, which is annoying
>>>>> and/or
>>>>> alarming to someone with no plans to do anything involving compilation of
>>>>> C
>>>>> code.
>>>>>
>>>>> The calls to xcode-select were added in PR#36599
>>>>> <https://github.com/sagemath/sage/pull/36599> in order to force XCode
>>>>> to use Apple's ld-classic linker instead of ld when their new version of
>>>>> ld
>>>>> was totally broken. This is done by adding -ld_classic to LDFLAGS.
>>>>>
>>>>> *Note to people who worked on PR #36599* (@jhpalmieri and @mkoeppe):
>>>>> I think Apple's new linker is working now, so it is probably no longer
>>>>> necessary and not a good idea to force use of ld_classic.
>>>>>
>>>>> - Marc
>>>>>
>>>>> On Tuesday, April 23, 2024 at 10:45:54 PM UTC-5 Marc Culler wrote:
>>>>>
>>>>> That was it!
>>>>>
>>>>> Thank you Gonzalo; indeed, it helps a lot. And your workaround is
>>>>> fine, since we don't support the -i option, Even if we did, the default
>>>>> names for as, ld and ar are correct whenever the command line tools are
>>>>> installed. So that block of code is completely irrelevant for the macOS
>>>>> platform. This solves the problem.
>>>>>
>>>>> - Marc
>>>>>
>>>>>
>>>>>
>>>>> On Tuesday, April 23, 2024 at 8:59:12 PM UTC-5 Gonzalo Tornaría wrote:
>>>>>
>>>>> https://github.com/sagemath/sage/blob/develop/src/bin/sage-env#L482
>>>>> and L494
>>>>>
>>>>> See: https://github.com/sagemath/sage/issues/14296 and
>>>>> https://github.com/sagemath/sage/commit/69213d74ead4e93687cf61f214b0d96dd3f9885a
>>>>>
>>>>> Maybe you can workaround this by setting AS=as and LD=ld in
>>>>> sage-env-config.
>>>>>
>>>>> HTH,
>>>>> Gonzalo
>>>>>
>>>>>
>>>>> On Tuesday, April 23, 2024 at 3:48:18 PM UTC-3 Marc Culler wrote:
>>>>>
>>>>> I discovered, by installing the Sage_macOS app on a pristine macOS
>>>>> system, that somehow, somewhere, in Sage's startup sequence there is a
>>>>> call
>>>>> to gcc. This is true whether Sage is being started from a command line
>>>>> or
>>>>> a notebook.
>>>>>
>>>>> On such a macOS system /usr/bin/gcc exists, but calling it causes a
>>>>> dialog to be posted which asks whether to download and install the Xcode
>>>>> "command line tools".
>>>>>
>>>>> There is no need for a user to install, or be prompted to install, a C
>>>>> compiler in order to run Sage. If we want to verify whether a C compiler
>>>>> is installed on the host system then we should check the return value of
>>>>> xcode-select
>>>>> -p rather than calling /usr/bin/gcc.
>>>>>
>>>>> I am unable to find where this call occurs. Do any of the Sage
>>>>> developers know which component of Sage could be calling /usr/bin/gcc on
>>>>> start up?
>>>>>
>>>>> - Marc
>>>>>
>>>>> --
>>>> You received this message because you are subscribed to a topic in the
>>>> Google Groups "sage-devel" group.
>>>> To unsubscribe from this topic, visit
>>>> https://groups.google.com/d/topic/sage-devel/4s_5HmznHZM/unsubscribe.
>>>> To unsubscribe from this group and all its topics, send an email to
>>>> sage-devel+...@googlegroups.com.
>>>> To view this discussion on the web visit
>>>> https://groups.google.com/d/msgid/sage-devel/c81c9434-3b0c-4730-b3e8-6e36802f5224n%40googlegroups.com
>>>>
>>>> <https://groups.google.com/d/msgid/sage-devel/c81c9434-3b0c-4730-b3e8-6e36802f5224n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>>
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/sage-devel/75de4cfb-75cf-4f86-9fbf-3b0de1bbeac7n%40googlegroups.com.
