Fix infinite recursion in E.econd when optimizing nested conditionals

[nojaf]

So, my project over at nojaf/rescript-kaplay@2f34e58 was hanging this morning.
bsc.exe was in an infinite loop.

The LLM and I chatted about this and we found that the problem was an infinite recursion in compiler/core/js_exp_make.ml.

I added the technical context in debug_compiler_hang.md.

@cristianoc could you take a look at this please?

I was not able to reproduce this outside of my own project.
So, this might be a quite specific edge-case 🤔

[pkg-pr-new]

Open in StackBlitz

rescript

npm i https://pkg.pr.new/rescript-lang/rescript@8039

@rescript/darwin-arm64

npm i https://pkg.pr.new/rescript-lang/rescript/@rescript/darwin-arm64@8039

@rescript/darwin-x64

npm i https://pkg.pr.new/rescript-lang/rescript/@rescript/darwin-x64@8039

@rescript/linux-arm64

npm i https://pkg.pr.new/rescript-lang/rescript/@rescript/linux-arm64@8039

@rescript/linux-x64

npm i https://pkg.pr.new/rescript-lang/rescript/@rescript/linux-x64@8039

@rescript/runtime

npm i https://pkg.pr.new/rescript-lang/rescript/@rescript/runtime@8039

@rescript/win32-x64

npm i https://pkg.pr.new/rescript-lang/rescript/@rescript/win32-x64@8039

commit: 04485c4

[cristianoc]

Where's the self contained repro of a few lines?

[cristianoc]

Can't review without a repro.

[nojaf]

Where's the self contained repro of a few lines?

While I understand that request, I can't reproduce it with the problematic code in a single file. I do not know what exactly trigger it, thus it must be some very unique combination.

git clone --no-checkout https://github.com/nojaf/rescript-kaplay.git
cd rescript-kaplay
git checkout 2f34e581f346bff016bcc99126ba9656565f7ca6
bun i 
bunx rescript

[cristianoc]

Where's the self contained repro of a few lines?

While I understand that request, I can't reproduce it with the problematic code in a single file. I do not know what exactly trigger it, thus it must be some very unique combination.

git clone --no-checkout https://github.com/nojaf/rescript-kaplay.git
cd rescript-kaplay
git checkout 2f34e581f346bff016bcc99126ba9656565f7ca6
bun i 
bunx rescript

Can you reproduce it on the full repo, and reproduce it going away after the fix in this PR?

[nojaf]

Yep, I tried to mention this in https://github.com/rescript-lang/rescript/pull/8039/files#diff-e4faf79cc44b60806dfcf3de513d06c907723189948ac3738b02d4939252f523R109-R128

Works for that change but not for v12 stable.

[cristianoc]

Yep, I tried to mention this in https://github.com/rescript-lang/rescript/pull/8039/files#diff-e4faf79cc44b60806dfcf3de513d06c907723189948ac3738b02d4939252f523R109-R128

Works for that change but not for v12 stable.

can you ask the agent to cut down your project little by little until what's left is small and still loops?

[cristianoc]

We want to make sure that we're not going after a red herring.

[nojaf]

I'm able to trim some more in https://github.com/nojaf/rescript-kaplay/tree/minimal-repro

It seems related to https://github.com/nojaf/rescript-kaplay/blob/minimal-repro/packages/skirmish/src/Player.res#L12

Thundershock.cast(gameObj) is the root cause of the loop.

[cristianoc]

Not an infinite loop: see repro
#8042

The original example is over-engineered, so it took some time to unentangle this.

[cristianoc]

See repro.
This is a perf issue not an infinite loop issue.
Ideally we want a solution that avoids the problem and does not change any of the current tests (unless there's a good reason to).
There are several ways to go about this, but prompting from a small repro should make this much easier.