Failure to compute Massey products <h3^2, d0, ->

[JoeyBF]

If I'm not mistaken, the product h3^2 * d0 is zero (the target bidegree is empty). However, trying to compute the Massey products gives me the following error:

$ cargo run --example massey -- S_2 "" "" "" 14 2 "[1]" 14 4 "[1]"
    Finished dev [unoptimized + debuginfo] target(s) in 0.02s
     Running `target/debug/examples/massey S_2 '' '' '' 14 2 '[1]' 14 4 '[1]'`
Module (default: S_2): S_2
Module save directory (optional):
Max n (default: 30):
Max s (default: 7):

Computing Massey products <a, b, ->

Enter a:
n of Ext class a: 14
s of Ext class a: 2
Input Ext class a: [1]

Enter b:
n of Ext class b: 14
s of Ext class b: 4
Input Ext class b: [1]
thread 'main' panicked at 'Failed to lift at (target_s, target_t) = (1, 3)', /home/joey/Documents/sseq/ext/src/chain_complex/chain_homotopy.rs:260:17
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[dalcde]

I don't have time to investigate at the moment, but I'll note that h3 d0 is already zero (if only there were a computer program that could tell us that...), but computing that does not panic.

The failure is in very low degree so I would guess a few prints would quickly pinpoint the issue.

[JoeyBF]

Yes, the same problem happens with a = h0^2. If b is killed by h0^2 but not h0, there's no panic. But if b is also killed by h0 itself, it panics

[dalcde]

Just to check my maths <h0², b, -> is just <h0, b, h0> multiplication (up to indeterminacy), right?

On 27 November 2022 10:30:38 pm GMT-05:00, Joey Beauvais-Feisthauer @.***> wrote:

Yes, the same problem happens with a = h0^2. If b is killed by h0^2 but not h0, there's no panic. But if b is also killed by h0 itself, it panics

-- Reply to this email directly or view it on GitHub: https://github.com/SpectralSequences/sseq/issues/116#issuecomment-1328489059 You are receiving this because you commented.

Message ID: @.***>

[JoeyBF]

My gut is telling me no but I can't think of a counterexample. You might be right.

BTW I think the bug is because it tries to lift homotopies when the product is not nullhomotopic. I just caught it trying to compute a homotopy for h1 * 1 (the identity in (0,0))