Installing troubles (on cluster)

[Atulsingh92]

Hello all,

Finding it real hard to install this on the cluster. ( as of 25/08/2025)

1. When the instructions says "Replace "1.6" by the current version number", where do I then find the common package of this current version.
2. Does the "current version" imply whatever is the stable version on git?, and then for the common, I download whichever is the latest in the list and rename to the version in point 1.?
   2a. Note that the common .tar file download page, then has tar file only for 1.17, 1.16, and 1.9.2, so can only install the corresponding Macaulays?
3. git clone instructions -> make ->make: *** No targets specified and no makefile found. Stop. -> SO -> Renamed Makefile.in to makefile -> make: *** No rule to make target 'include/config.Makefile'. Stop. Why the several levels of complexity, or am I missing something?
4. Again, is there a difference in 1.6 and 1.24? (as the files on the list seems to be from 1.9 to 1.24)
5. Most importantly, if i386 is to be replaced with the architecture x86_64 (say) ? And it will work?
6. Macaulay2-1.2/INSTALL does not appear after tar xzf Macaulay2-whatever !!!, why is that? (Also where is that, as there are now several links that I have opened and cannot back to "this website" where the instructions suggest to find the INSTALL file. !!

[d-torrance]

Apologies for the confusion! In general, the instructions on macaulay2.com are quite out-of-date. The best place to get installation instructions is the wiki: https://github.com/Macaulay2/M2/wiki.

As for your specific questions:

1. We haven't published common .tar files in quite a few releases. The recommended way to install Macaulay2 is either using a pre-built package (Debian, homebrew, etc.) or building from source after cloning the git repository
2. Building the stable branch in git is the way to go. It's the latest release (currently, 1.25.06), with possibly some additional patches to fix build issues that have cropped up.
3. The first run of make in the autotools build was replaced with the more standard ./autogen.sh several releases ago.
4. Yes, 1.24 is much newer than 1.6. But 1.25.06 is the most recent -- build that instead. :)
5. Yes, i386 builds should be supported. You shouldn't need to specify anything -- the architecture is detected during the initial configuration step.
6. That file was removed and its content moved to the wiki some time ago.

[mahrud]

Also, since you're building on a cluster you may want to manually raise maxNumThreads here:

M2/M2/Macaulay2/system/supervisor.cpp

Lines 15 to 16 in 2cb101f

	// We allocate between 5 and 17 threads initially, to save trouble with memory allocation.
	const static int maxNumThreads = ((numCores < 4) ? 4 : (16 < numCores ? 16 : numCores)) + 1;

[Atulsingh92]

Following this,
make still gives me make: *** No targets specified and no makefile found. Stop.

[d-torrance]

Did configure finish successfully? It's what generates the Makefile

[Atulsingh92]

I can run make for my other things.
Also,
the following is an example of what I see,

checking <something> <yes, no, none required>

Last two lines are

checking for tbb version... 2018.0
configure: tbb >= 2020 required; will build
configure: error: cmake is required to build tbb

Not sure if this counts as success or not?

[d-torrance]

No, configure didn't finish, so there's no Makefile yet.

You can either install cmake on the system, in which case the Macaulay2 build system will use cmake to build TBB, or add the --disable-tbb option to configure.

[Atulsingh92]

checking for boostlib >= 1.65 (106500)... yes
checking whether the Boost::Stacktrace library is available... no, will use header-only library
checking whether the Boost::Regex library is available... yes
configure: error: Could not find a version of the Boost::Regex library!

2. Do I need to address checking <something> <no> in the configure step?

3. Is there an example of successfull run of configure? Or it just runs without the last line error?

4. Tried with --disable-tbb, and the error is same as point 1.

(just an FYI, I did run the sudo dnf install <several packages> step as well from the instructions)

[d-torrance]

Yes, any time there's an error in the configure step, there's some troubleshooting to do. :)

There are lots of examples of successful runs of configure on the GitHub Actions page.

It's interesting that it says that Boost::Regex is available but that it couldn't find it. There should be a file named config.log that's generated by the configure script that gives more info about what exactly happened.