Compiling in Cygwin on Windows in 2025

[ofloveandhate]

So, it's 2025, and compiling is still a challenge. This PR is about compiling in Cygwin in 2025. It doesn't work as described in the user manual, sadly. A student wrote that section in ~2016 and I don't use windows, so I'm not surprised it's out of date. Let's work on it!

[hillscarol]

The issue I had with compiling in windows is presented here:

Bertini1 (BertiniSource_v1.6) compiles with some warnings, and libbertini-parallel files are present, but bertini_real does not recognize where these files are and aborts the configuration. Is there a special command I need to include in my PATH or my Bertini compilation so that it is recognized as a parallel compilation? I am following your compilation instructions for windows in your user manual (attached). I used the most recent cygwin package versions, but had to uninstall/reinstall libmpfr6 to libmpfr4 via cygwin64 in order to compile Bertini1.

For reference, additional software package downloads are:

• gpm-6.3.0
• mpfr-4.2.2
• mpc-1.3.1

Update: bertini_real compiled! My bertini-parallel.* and libbertini-parallel- files were not copied to the same places as bertini-serial. I did a dirty trick of copying the needed parallel files into the same locations as the serial ones. Then one can run ./configure as usual. Keep the thread open as there are still some issues persisting for me in calling make

[ofloveandhate]

I'm glad you want to use Bertini_real, it's an adventure! I haven't used it on Windows in many years, the windows instructions date to about 2016, so they almost certainly need updating.

You probably do need to tell Bertini_real where to find Bertini. Do you know where Bertini1 got installed to? /usr/local is the default location, but I don't know what happens to that location under Windows these days.

Here's a line I use to configure Bertini_real on my mac using dependencies installed by Homebrew (which end up in a location not scanned by the current build system):

./configure CPPFLAGS=-I/opt/homebrew/include LDFLAGS=-L/opt/homebrew/lib

The most relevant flags you probably need are CPPFLAGS and LDFLAGS, to specify the locations of Bertini1's headers and libraries respectively. I bet (but do not know) that if you can just figure out where b1 got installed, then you can pass those locations like I have above and things work out.

[ofloveandhate]

Update: bertini_real compiled! My bertini-parallel.* and libbertini-parallel- files were not copied to the same places as bertini-serial. I did a dirty trick of copying the needed parallel files into the same locations as the serial ones. Then one can run ./configure as usual. Keep the thread open as there are still some issues persisting for me in calling make

Fascinating, i'm suprised that the parallel things went somewhere else. Hmmmmmmmm.

[ofloveandhate]

What issues are you experiencing in calling make? Probably some compiler errors that are simple-ish to fix. Let me know so I can help!

[hillscarol]

The bertini-parallel files definitely got removed on a make clean call I did to try compiling again. Then when I called make again, they were not easily found, hence my copy & paste trick.

For the make call, I think I actually have everything compiled (I also am working on setting this up after work so my availability to deal with this installation is a bit spotty and I forget what I had previously faced). Below are screenshots of what my console outputs when trying to call bertini_real after calling bertini on the Cayley Cubic testrun.

[ESC309962:01622] PMIX ERROR: INIT in file /cygdrive/d/a/scallywag/openmpi/openmpi-4.1.6-1.x86_64/src/openmpi-4.1.6/opal/mca/pmix/pmix3x/pmix/src/mca/gds/ds21/gds_ds21_lock_pthread.c at line 187
[ESC309962:01622] PMIX ERROR: SUCCESS in file /cygdrive/d/a/scallywag/openmpi/openmpi-4.1.6-1.x86_64/src/openmpi-4.1.6/opal/mca/pmix/pmix3x/pmix/src/mca/common/dstore/dstore_base.c at line 2454
[ESC309962:01622] PMIX ERROR: INIT in file /cygdrive/d/a/scallywag/openmpi/openmpi-4.1.6-1.x86_64/src/openmpi-4.1.6/opal/mca/pmix/pmix3x/pmix/src/mca/gds/ds12/gds_ds12_lock_pthread.c at line 138
[ESC309962:01622] PMIX ERROR: SUCCESS in file /cygdrive/d/a/scallywag/openmpi/openmpi-4.1.6-1.x86_64/src/openmpi-4.1.6/opal/mca/pmix/pmix3x/pmix/src/mca/common/dstore/dstore_base.c at line 2454
[ESC309962:01621] PMIX ERROR: ERROR in file /cygdrive/d/a/scallywag/openmpi/openmpi-4.1.6-1.x86_64/src/openmpi-4.1.6/opal/mca/pmix/pmix3x/pmix/src/mca/gds/ds12/gds_ds12_lock_pthread.c at line 168

Then I get the expected output for stabilization / deflation. My issue lies when python is called to create the critical system. There is an issue in my cygwin installation of numpy with the python package, the classic "no module named numpy".

File "/cygdrive/c/cygwin64/usr/local/hillsc/bertini_real/test/init_test/python_nul
import numpy as np
ModuleNotFoundError: No module named 'numpy'

Trying to install numpy via pip does not work as the installation fails on the subprocess to install backend dependencies, including ninja.

[ofloveandhate]

I'm glad you were able to get make to work. On the topic of compiling, I did just implement a CMake build system for Bertini_real, so the pattern of

mkdir build && cd build
cmake ../
make 
make install

should work now, and be more reliable than the Autotools based system that was in place. (I left autotools for now, but i'll fully remove it some day).

This build system improvement does not solve the OpenMPI or numpy issues, obviously

[ofloveandhate]

I have not forgotten about this! I have ordered a Windows computer!!! I hope to refresh the Cygwin instructions after the new machine arrives.