[WIP] Update Symbolics and SymbolicUtils version

[0x0f0f0f]

NOTE: Symbolics 4.0.0 has not yet been released.

[0x0f0f0f]

Tests are all green locally with latest commit from JuliaSymbolics/Symbolics.jl#421
Will notify here when Symbolics.jl 4.0.0 gets registered.

[0x0f0f0f]

That segfault in StructuralIdentifiability.jl looks like something I'm not able to fix myself, I'll need some help. @shashi @ChrisRackauckas

[ChrisRackauckas]

Can you isolate the issue a bit?

[0x0f0f0f]

Can you isolate the issue a bit?

Testing locally.

[0x0f0f0f]

Can you isolate the issue a bit?

Cannot replicate on my machine.
StructuralIdentifiability.jl tests pass locally on my machine using this commit of MTK, latest MTK tag and StructuralIdentifiability.jl on latest master branch. Seems like the issue is due to libgmp in CI container.

 signal (11): Segmentation fault
in expression starting at /home/runner/work/ModelingToolkit.jl/ModelingToolkit.jl/downstream/test/identifiability.jl:1
__gmpn_addmul_1_coreisbr at /opt/hostedtoolcache/julia/1.6.3/x64/bin/../lib/julia/libgmp.so (unknown line)
Allocations: 196265728 (Pool: 196227157; Big: 38571); GC: 102
ERROR: LoadError: Package StructuralIdentifiability errored during testing (received signal: 11)

https://github.com/SciML/ModelingToolkit.jl/runs/4073261736?check_suite_focus=true

[ChrisRackauckas]

What's the function call that causes a segfault?

[0x0f0f0f]

What's the function call that causes a segfault?

Doesn't show in the logs, only shows Pkg testing errors. Cannot find out since I cannot replicate locally
https://github.com/SciML/ModelingToolkit.jl/runs/4073261736?check_suite_focus=true#step:6:653

Only reports function __gmpn_addmul_1_coreisbr from libgmp

[ChrisRackauckas]

What is the thing right before it in the tests, and after?

[0x0f0f0f]

What is the thing right before it in the tests, and after?

Right before:

[ Info: Assessing global identifiability using the coefficients of the io-equations

Third time that this line appears in the log. There's nothing after.

[ChrisRackauckas]

No, what piece of code is ran that generates the segfault?

[0x0f0f0f]

What is the thing right before it in the tests, and after?

Judging from where the logs stop, the segfault should be happening here
https://github.com/SciML/StructuralIdentifiability.jl/blob/5dade8f547b3c68d527c7e67197f49a17bfa1948/src/global_identifiability.jl#L224-L225

[ChrisRackauckas]

Isolate that to an MWE. I think it must be due to the conversion of the IR.

[0x0f0f0f]

Isolate that to an MWE. I think it must be due to the conversion of the IR.

Any suggestions on how to proceed knowing that the issue does not subsist on my local dev environment? I'll try replicating the downstream test in a docker container with https://github.com/nektos/act

[ChrisRackauckas]

So when you run the tests it works?

[0x0f0f0f]

So when you run the tests it works?

Yes. Running on void linux. Using glibc. gmp v6.2.1_1 and gmpxx v6.2.1_1. Julia v1.6.1.

[ChrisRackauckas]

Newer Julia? v1.6.3 is what the tests will use. Double check if the dependencies in the Symbolics stack are all similar.

[0x0f0f0f]

@ChrisRackauckas passes locally on v1.6.3. Testing github action in docker container

[0x0f0f0f]

Passes locally when running the github action in a docker container!!! Can you try retriggering the workflow?

[0x0f0f0f]

Dependency version differences from the Github CI and the GH action test in docker container on my local machine are:
ArrayInterface v3.1.38 => v3.1.39
Parsers v2.1.1 => Parsers v2.1.2
VectorizationBase v0.21.19 => v0.21.21

[0x0f0f0f]

Dependency version differences from the Github CI and the GH action test in docker container on my local machine are: ArrayInterface v3.1.38 => v3.1.39 Parsers v2.1.1 => Parsers v2.1.2 VectorizationBase v0.21.19 => v0.21.21

Test passed. probably related to one of those packages. Should bump compat? @ChrisRackauckas

[ChrisRackauckas]

Yeah bump the lower bound to avoid the issue.

[0x0f0f0f]

lgtm? @ChrisRackauckas

[shashi]

We should make a breaking release of MTK after this, so that further downstream updates can be more cautious.