Question about FiniteFlow exp branch

[aweaver1fandm]

I will begin by saying that I am aware that you may not be able to answer the question I'm about to ask.

In relation to issue #6 I did install the exp branch of FiniteFlow and made sure to build it with -DFFLOW_USE_FLINT=0.
I am building FiniteFlow for a researcher here who needs it for use with AMFlow (https://gitlab.com/multiloop-pku/amflow). As a test, he tried the following script to evaluate a single 3-loop integral. This script works on his laptop which is admittedly running the master branch of FiniteFlow

xValue=1/2;
threads=6;
Get["/opt/amflow/AMFlow.m"]
AMFInitialize["FiniteFlow+LiteRed"];
SetReductionOptions["IBPReducer" -> "FiniteFlow+LiteRed", "BlackBoxRank" -> 4, "BlackBoxDot" -> 3];
AMFlowInfo["Family"] = G5;
AMFlowInfo["Loop"] = {q, r, s};
AMFlowInfo["Leg"] = {p};
AMFlowInfo["Conservation"] = {};
AMFlowInfo["Replacement"] = {p^2 -> 1};
AMFlowInfo["Propagator"] = {q^2, r^2, s^2, q^2 - 2 x q p,  r^2 - 2 x r p, s^2 + 2 s p, (q - s)^2 - 2 x (q p - s p), (r - s)^2 -  2 x (r p - s p), (q + r - s)^2 - 2 x (q p + r p - s p)};
AMFlowInfo["Numeric"] = {x -> xValue};
AMFlowInfo["NThread"] = threads;
integrals =
   {j[G5, 1, 0, 1, 0, 1, 1, 1, 0, 1]};
solG5A1 = SolveIntegrals[integrals, 42, 12]
Save["/home/gadkins/amf5/solG5A1", solG5A1];

When he runs it with the exp branch of FiniteFlow, the script fails with the following error

FF::logerr: New set of needed variables is not a subset of the previous one
"number of independent eqs." -> 404151

Is there something you might be aware of that would account for this error? Is there something that we could change in the script to make it work? I also understand that AMFlow itself may need to be updated in order to more closely match what you're doing with FiniteFlow. That's part of why I'm asking, to figure out how to properly proceed.

Thank you

[peraro]

Hi, thank you for reporting this.

I am not deeply familiar with the internals of AMFlow, but my colleagues have successfully used it with the exp branch.

I have two questions, which may help identifying the issue:

1. Are you sure you are on the latest commit of exp (earlier versions were known to have issues)?
2. Can you confirm if the AMFlow evaluation is really failing, or if you are just assuming that because an error is being printed? Note that this is a FiniteFlow error, and I would expect a failed evaluation to also print an AMFlow error message. Other users have reported the same error being printed, while AMFlow still continued normally. I suspect that the same error may appear in master as well, except that FiniteFlow wouldn't print it. This error does not immediately prevent an application from continuing, it just signals that restricting the number of needed variables did not work. It might still fail at a later stage, if this was critical.

[aweaver1fandm]

1. I am certain we are using the most up to date exp. Git pull says it's already up to date and git log shows the following at the beginning of the log

commit 1cf05c56eda178c0d51b15c0faed326b532935d0 (HEAD -> exp, origin/exp)
Author: Tiziano Peraro <tiziano.peraro@unibo.it>
Date:   Fri Jun 5 14:41:34 2026 +0200

    Fix detection of impossible linear fits, in Mathematica interface.

2. Yes it truly does error out. Here is the output from the user's run.

AMFSystemSetup: setting up system 1.
Family -> G5
Loop -> {q, r, s}
Leg -> {p}
Conservation -> {}
                 2
Replacement -> {p  -> 1}
                2   2   2   2             2                            2         2                           2                               2
Propagator -> {q , r , s , q  - 2 p q x, r  - 2 p r x, -eta + 2 p s + s , (q - s)  - 2 (p q - p s) x, (r - s)  - 2 (p r - p s) x, (q + r - s)  - 2 (p q + p r - p s) x}
                 1
Numeric -> {x -> -}
                 2

AMFSystemDifferentialEquation: preparing differential equations for system 1.
RunProcess failed for 1 time:  -> {{/usr/local/Wolfram/Wolfram/14.3/SystemFiles/Kernel/Binaries/Linux-x86-64/WolframKernel, -noprompt, -script, 
>     /home/gadkins/amf5/cache/G5_amflow/1/diffeqsetup/reductionDe.wl}, Inherited}

Time wasted:  -> 136.825

RunProcess failed for 2 times:  -> {{/usr/local/Wolfram/Wolfram/14.3/SystemFiles/Kernel/Binaries/Linux-x86-64/WolframKernel, -noprompt, -script, 
>     /home/gadkins/amf5/cache/G5_amflow/1/diffeqsetup/reductionDe.wl}, Inherited}

Time wasted:  -> 278.259

RunProcess failed for 3 times:  -> {{/usr/local/Wolfram/Wolfram/14.3/SystemFiles/Kernel/Binaries/Linux-x86-64/WolframKernel, -noprompt, -script, 
>     /home/gadkins/amf5/cache/G5_amflow/1/diffeqsetup/reductionDe.wl}, Inherited}

Time wasted:  -> 421.686

RunCommand: process failed -> {/usr/local/Wolfram/Wolfram/14.3/SystemFiles/Kernel/Binaries/Linux-x86-64/WolframKernel, -noprompt, -script, 

>    /home/gadkins/amf5/cache/G5_amflow/1/diffeqsetup/reductionDe.wl}
Missing[KeyAbsent, StandardError]

Out[1]= $Aborted

They are using the same script they use on their laptop so I am (perhaps erroneously) assuming that it's not a typo in the script or something like that

[peraro]

I had a look with the help of two collaborators. They tested AMFlow using the same input you quoted above. While they were been able to reproduce the FiniteFlow error message, AMFlow still continued without issues for them, so I am not sure the message is related to your issue.

To fix FiniteFlow's error message, you could try applying the patch attached using git apply issue11_patch1.patch. It turns out ResetNeededVars in the master branch was a bit more lenient than in the exp branch. This was mostly an accident, due to internal details which have since changed, but amflow now seems to rely on it. The patch should make exp backward compatible with master. However, I am not sure it will resolve your issue, but let me know in case it does.

issue11_patch1.patch

[aweaver1fandm]

I will try to apply the patch and see what happens. Since it does seem to work for your collaborators, could part of the issue be 1. How I built FiniteFlow? I only included the

…

-DCMAKE_INSTALL_PREFIX=/installation/path/prefix and Flint = 0 options. and/or 2. A Mathematica version issue?

On Tue, Jun 16, 2026 at 11:04 AM peraro ***@***.***> wrote: *peraro* left a comment (peraro/finiteflow#11) <#11 (comment)> I had a look with the help of two collaborators. They tested AMFlow using the same input you quoted above. While they were been able to reproduce the FiniteFlow error message, AMFlow still continued without issues for them, so I am not sure the message is related to your issue. To fix FiniteFlow's error message, you could try applying the patch attached using git apply issue11_patch1.patch. It turns out ResetNeededVars in the master branch was a bit more lenient than in the exp branch. This was mostly an accident, due to internal details which have since changed, but amflow now seems to rely on it. The patch should make exp backward compatible with master. However, I am not sure it will resolve your issue. issue11_patch1.patch <https://github.com/user-attachments/files/29007416/issue11_patch1.patch> — Reply to this email directly, view it on GitHub <#11?email_source=notifications&email_token=ANJXBCYGKMOKKZ6VJD4ZD635AFOXRA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZSGAZDGMJYHAZKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-4720231882>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ANJXBC2RTXJE7KSFJHBOMNT5AFOXRAVCNFSNUABFKJSXA33TNF2G64TZHMYTQNZXGEYDGNBXHNEXG43VMU5TINRWGQ4DQMRRGM2KC5QC> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ANJXBC4LBZXHU5TCW4OTF4L5AFOXRA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZSGAZDGMJYHAZKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/ANJXBCZZD55Y6TE2HAZMAED5AFOXRA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZSGAZDGMJYHAZKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[aweaver1fandm]

I applied the patch, rebuilt finiteflow and it still aborts. I made sure to clear out cache just to make sure but still no luck. Is there extra commands I can add to try to get more verbose output or print statements or something to try to narrow down what is going on?

Thank you

[peraro]

They also tried it without Flint, but the outcome was the same.

Do you have any log file of AMFlow with additional details? The error messages above aren't specific enough to understand why it is failing for you.

Based on the information you provided and a quick search of AMFlow's code, my guess is that the script reductionDe.wl generated here is failing (although, I am not familiar enough with AMFlow's code to be sure).

I would check if there is any further useful output from AMFlow (I understand this info is sometimes buried somewhere in the cache directory) that I could look at. Alternatively, you could add some Print statements inside BasicCode[], CheckCode[] and extra defined in ibp_interface/FiniteFlow+LiteRed/interface.m to see where it is failing and get a better idea.

[peraro]

Other users reported crashes with the exp branch, and it turned out they were loading an old FiniteFlow.m file (which happened to be in their $Path) instead of the one in exp.

This can be checked, for example, by evaluating $FFVersion after loading FiniteFlow in Mathematica, which should evaluate to 2 if the exp version is being loaded. In your case, you should check this by adding anything amflow also adds to the $Path.

Since my collaborators could not replicate your error, there's a chance you might have an issue similar to the one above.

If you could also test any other script, such as mathlink/tests.m and mathlink/test_sparse_solver.m, they could give useful information as well.