✔️ 99ea46c -> Azure artifacts URL

✔️ 7b542f7 -> Azure artifacts URL

Codecov Report

Merging #2484 (7b542f7) into master (d3b7e7b) will decrease coverage by 0.08%.
The diff coverage is n/a.

❗ Current head 7b542f7 differs from pull request most recent head ae688ab. Consider uploading reports for the commit ae688ab to get more accurate results

@@            Coverage Diff             @@
##           master    #2484      +/-   ##
==========================================
- Coverage   61.16%   61.08%   -0.08%     
==========================================
  Files         630      627       -3     
  Lines      121443   121064     -379     
==========================================
- Hits        74276    73950     -326     
+ Misses      47167    47114      -53     

see 13 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

To fix #2399, we also need to make it work on mac. (Not sure if it works on x86_64, but it definitely doesn't work on M2).

To fix #2399, we also need to make it work on mac. (Not sure if it works on x86_64, but it definitely doesn't work on M2).

What does not work on M2? The use of ncurses-6.4? Or the existing wheel for 8.2.2 with, presumably, ncurses-6.2. I'm trying to build a wheel on my apple M1 but have not succeeded yet.

The reason I though it was unique to linux wheels is that I don't see the backspace problem on my M1 with 8.2.2.

The wheel for 8.2.2 with ncurses 6.2 does not work on a Mac.

If everything works with ncurses 6.4 that seems better than reverting to an old version.

The reason I though it was unique to linux wheels is that I don't see the backspace problem on my M1 with 8.2.2.

That's fascinating, because I have the issue on my M2, and Pramod's #2399 shows the issue on an x86_64 mac.

Now I see the issue on my M1. Perhaps I had the wrong environment.

I'm still having trouble building a wheel on the M1, though now I'm getting almost to the end...

 => LINKING shared library ./libnrnmech.dylib
 => LINKING executable ./special LDFLAGS are:       
Successfully created arm64/special
dyld[29515]: symbol not found in flat namespace '_BC'
/Users/hines/neuron/backspace/build/cmake_install/bin/neurondemo: line 34: 29515 Abort trap: 6           ${NRNBIN}nrniv -dll "${NRNDEMO}release/${MODSUBDIR}/.libs/libnrnmech.so" "${NRNDEMO}demo.hoc" "$@" -

I'm trying with

bash packaging/python/build_wheels.bash osx 3.11

@nrnhines : Sorry, it was my ignorance! I see that we also use static readline + ncurses 6.2 for Mac OS: https://github.com/neuronsimulator/nrn/blob/master/packaging/python/build_static_readline_osx.bash.

Will get that updated soon!

I figured out that _BC is undefined in libreadline :)

@pramodk Great! With your hint, I successfully built an osx wheel for my M1 using ncurses-6.4 and it does work.

Question: I presume this PR fixes the issue on osx for the next neuron-nightly. But the linux issue does not get fixed until a new Docker is published. Then our normal CI is used to build the linux neuron-nightly. I guess I'm a bit shaky on how to get all this put back into 8.2.2 (or 8.2.3?)

But the linux issue does not get fixed until a new Docker is published.

I have pushed the updated image to DockerHub. So that should be also good!

I successfully built an osx wheel for my M1 using ncurses-6.4 and it does work.

Tested the same on x86 Mac and it works for me as well.

I guess I'm a bit shaky on how to get all this put back into 8.2.2 (or 8.2.3?)

I guess we should be able to rebuild older releases as you hinted. Need to think a bit about what's the "easier" way to do this. But I guess it's not urgent and can happen in the coming week(s)?

Ted wrote

Strange stuff also happens under Win 10.

Start bash, launch nrngui and start executing statements, and soon the cursor disappears, and cursor images randomly "stick" to lines that have already been executed. Backspace seems to work properly.

Instead, just open an MSWin command prompt, execute nrngui, and start executing statements. Similar cursor problems appear, and backspace doesn't work correctly.

I notice in ci/win_install_deps.cmd the line

mingw-w64-x86_64-ncurses ^

I wonder if 8.2.2 was getting ncurses-6.2? (And how could we know?)

I wonder if 8.2.2 was getting ncurses-6.2? (And how could we know?)

Do we have this "weird" behaviour only with 8.2 on Windows and older version is working fine?

I guess I'm a bit shaky on how to get all this put back into 8.2.2 (or 8.2.3?)

Needs to be 8.2.3. The installation has changed. Features are different (e.g. backspace exists). Bugfixes are what minor versions are for. Silent updates don't get fixed on centrally managed systems and cause confusion. (e.g., "This works on my machine but not yours, and we're all using exactly the same software.")

But I guess it's not urgent and can happen in the coming week(s)?

HOC users would disagree. I have one specific person in mind.

✔️ ae688ab -> Azure artifacts URL

Logfiles from GitLab pipeline #149119 (:no_entry:) have been uploaded here!

Status and direct links:

• ✅ mac_m1_cmake_build: [cmake, ON, OFF, OFF, address]
• ✅ spack_setup
• ✅ build:nmodl
• ✅ build:neuron:nmodl:intel:legacy
• ✅ build:neuron:nmodl:intel:shared
• ✅ build:neuron:nmodl:nvhpc:acc:legacy
• ✅ build:neuron:nmodl:nvhpc:acc:shared
• ✅ build:neuron:nmodl:nvhpc:omp:legacy
• ✅ build:neuron:nmodl:nvhpc:omp
• ✅ test:neuron:nmodl:intel:legacy
• ✅ test:neuron:nmodl:intel:shared
• ✅ test:neuron:nmodl:nvhpc:acc:legacy
• ⛔ test:neuron:nmodl:nvhpc:acc:shared
• ✅ test:neuron:nmodl:nvhpc:omp:legacy
• ✅ test:neuron:nmodl:nvhpc:omp

330/358 Test #126: coreneuron_modtests::spikes_mpi_file_mode_py_gpu .........................***Failed  125.20 sec
srun: error: Unable to confirm allocation for job 1699382: Socket timed out on send/recv operation
srun: Check SLURM_JOB_ID environment variable. Expired or invalid job 1699382

one specific test that failed for GPU is unrelated, cluster error. So I am going to ignore it.

I downloaded the wheels produced from CI and they worked on Mac OS x86_64 and Linux. So I will merge this PR.

I wonder if 8.2.2 was getting ncurses-6.2? (And how could we know?)

@nrnhines : for windows, looking at the GitHub build log here, following versions of readline & ncurses are used today:

2023-09-01T15:25:29.7678358Z  mingw-w64-x86_64-readline-8.2.001-6-any downloading...
2023-09-01T15:25:29.7670430Z  mingw-w64-x86_64-ncurses-6.4.20230708-1-any downloading...

So at least ncurses/readline versions are now newer.

At the time of 8.2, difficult to know now. But considering pacman packages are relatively newer (?), probably there were no old versions.

Does the nightly windows installer work fine?

Like Mac/Linux, we might need to debug the windows issues separately. So I will go ahead and merge this.