Bumped ubuntu builds to 24.04 #3288
Conversation
3fc60e2
to
b80941d
Compare
|
Not sure what to do about the |
|
I think the following should fix the autotools-ubuntu build: --- a/.github/workflows/test_build.yml
+++ b/.github/workflows/test_build.yml
@@ -198,7 +198,8 @@ jobs:
- name: Validate HTML documentation
if: matrix.build-system == 'autotools' && runner.os == 'Linux'
run: |
- python3 -m pip install html5validator
+ sudo apt install pipx
+ pipx install html5validator
if test ${{ github.event_name }} = pull_request
then
git fetch --depth=1 origin development |
|
@mikestillman if you have an M1 machine, could you investigate why this is happening? -- capturing example results for "Lex" /bin/sh: line 1: 18364 Trace/BPT trap: 5 /Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/arm64-Darwin-macOS-14.5/bin/M2 -q --stop --silent --no-preload --no-threads -e errorDepth=3 -e debugLevel=0 -e "installPackage(\"Macaulay2Doc\", Verbose => false, RerunExamples => false, CheckDocumentation => true, IgnoreExampleErrors => false, RemakeAllDocumentation => false, InstallPrefix => \"/Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/\", UserMode => false, SeparateExec => true, DebuggingMode => true)" -e "\"info-\"|\"Macaulay2Doc\" << Macaulay2Doc#\"test number\" << close" -e "exit 0"Googling |
Without |
cc37c56
to
db9463d
Compare
|
@mikestillman could you try running the "Lex" example on arm64 with macos-14? There's no error file saved other than: -- capturing example results for "Lex" /bin/sh: line 1: 18412 Trace/BPT trap: 5 |
|
On CMake (macos-14 arm64), we're getting a core dump on this line: i5 : C = QQ[a..d,MonomialOrder=>Lex,MonomialSize=>8];
i6 : try a^(2^15-1) else "failed"(GRevLex is also causing the same crash) Meanwhile, autotools (macos-13 x86_64) is failing on several ForeignFunctions examples: i1 : mpfr = openSharedLibrary "mpfr"
stdio:1:8:(3): error: dlopen(libmpfr.dylib, 0x0001): tried: '/Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/x86_64-Darwin-macOS-13.6.7/bin/../lib/Macaulay2/lib/libmpfr.dylib' (no such file), '/libmpfr.dylib' (no such file), 'libmpfr.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OSlibmpfr.dylib' (no such file), '/Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/x86_64-Darwin-macOS-13.6.7/lib/Macaulay2/lib/libmpfr.dylib' (no such file), '/Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/x86_64-Darwin-macOS-13.6.7/lib/libmpfr.dylib' (no such file), '/usr/local/opt/flint/lib/libmpfr.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OS/usr/local/opt/flint/lib/libmpfr.dylib' (no such file), '/usr/local/opt/gmp/lib/libmpfr.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OS/usr/local/opt/gmp/lib/libmpfr.dylib' (no such file), '/Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/x86_64-Darwin-macOS-13.6.7/lib/Macaulay2/lib/libmpfr.dylib' (no such file), '/Users/runner/work/M2/M2/M2/BUILD/build/usr-dist/x86_64-Darwin-macOS-13.6.7/lib/libmpfr.dylib' (no such file), '/usr/local/opt/flint/lib/libmpfr.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OS/usr/local/opt/flint/lib/libmpfr.dylib' (no such file), '/usr/local/opt/gmp/lib/libmpfr.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OS/usr/local/opt/gmp/lib/libmpfr.dylib' (no such file), '/usr/lib/libmpfr.dylib' (no such file, not in dyld cache), 'libmpfr.dylib' (no such file) |
|
Why isn't |
|
I ran a little experiment, and in macos-12, |
|
Oops, that messed up some |
722815d
to
db9463d
Compare
|
Just pushed a commit with a slightly different approach, adding |
|
Wait, there are still some |
|
Yeah, sorry that's not you. I thought adding Also, I'm against hardcoding paths in the wrapper (which is kinda useless these days anyway since we don't distribute other libraries). We could use the value of |
|
Fair point -- the wrapper that doesn't really do anything on most installations has always bugged me too. I'll go back to the previous solution. |
|
To be clear, I'm against hardcoding paths anywhere :) I'm wondering why the section of ForeignFunctions that deals with macOS didn't add the brew path correctly. Can you do an experiment to see if macos-13 has a different value for any of the parameters? |
|
The macOS hack in ForeignFunctions currently only fires on arm machines, since |
d518b51
to
d976f4f
Compare
|
There was definitely a build were the linking issues didn't happen: https://github.com/Macaulay2/M2/actions/runs/9569457804/job/26382031099 |
|
@mikestillman have you had a chance to build M2 with clang on macos-14? I noticed the linked libraries of the binary built on github are strange: For instance, there are two different |
|
I'm generally against removing examples/tests because this is pointing to a regression that should be fixed rather than suppressed, but I don't have very strong feelings about it. |
|
I agree that we should definitely avoid skipping the test if at all possible. One big difference I see is that we're using factory from homebrew instead of building it. From the logs from this PR: But from a recent successful build: |
|
Still failing!! |
|
Ahh, I just realized: previously we weren't running tests on macos at all! I'll revert to that for now. |
Starting macOS 13 (Ventura), install-info is missing from the system, which causes the autotools build to fail.
It was previously only a problem on ARM machines, since dlopen never checked in /opt/homebrew/lib. But it's apparently a problem on Intel Ventura machines as well, since after to switching to macos-13 for the GitHub builds, dlopen doesn't check in /usr/local/lib either.
86b970c
to
1cf8a79
Compare
|
I also reduced daily tests to weekly, removed the |
1cf8a79
to
f468233
Compare
f468233
to
36cece2
Compare
|
This PR is legitimately cursed ... |
|
I decided to just comment out that test for now, but I think I'll make a list of all these tests and examples disabled throughout the years so someone can look into them. |
72c5b6a
to
ee6dc0c
Compare
No description provided.