Sebastian Thiel

It's very likely that using `gitoxide` would be preferred here. It has its own security model which makes the use of untrusted configuration impossible, and I'd be inclined to say...

That sounds like a start. More importantly, I think the weaknesses of the `safe` parameter should be very well documented.

Even though I haven't read (or understood) the entire issue yet, I thought it's a good idea to ping @oyvindln in case they have seen this already.

Thanks for reporting! My guess is that the lack of determinism must have to do with the parallelism which is indeed configured in the tests as well. My hypothesis is...

> Is the idea that this is a test bug? Shouldn't tests that are running in parallel and accessing files, at least if writing to files, be writing to files...

Thanks so much for this *extremely* thorough investigation. To speed up my own processing of it, I ran it through a local Gemma 2.5 for summarisation, and then looked at...

> 1. (My worry is that it may occur more readily under circumstances that are not tested, in which case it might represent an as-yet unidentified nontrivial symlink traversal bug...

> I'm curious: would you expect setting a `thread_limit` of `1` explicitly (instead of [implicitly](https://github.com/GitoxideLabs/gitoxide/blob/c3f06ae424ab4e1918a364cabe8276297465a73a/gix-features/src/parallel/mod.rs#L115-L119)) to have the same effect on this test as disabling `gix-features/parallel`? Or is there something...

Wonderful, then I am extra-curious how this will turn out!

Thanks for the summary, setting the thread_limit to 1 should do the trick. ~~I think what I'd do next is to keep the tempdirectory of the test, and compare the...