builtins.match regex support "broken" on aarch64-darwin

[andir]

Describe the bug

I just tried to use poetry2nix on aarch64-darwin and the eval failed invalid regular expression. Apparently this is an issue with the aarch64-darwin build of Nix. I'm using Nixpkgs from NixOS/nixpkgs#105026.

The problem appears to only exist on aarch64-darwin and not on any of the platforms. Also nixUnstable vs stable nix doesn't make a difference.

aarch64-darwin

[aarch64-darwin] $ nix repl
Welcome to Nix version 2.3.10. Type :? for help.

nix-repl> builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
error: invalid regular expression '^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)', at (string):1:1

[aarch64-darwin] $ nix repl
Welcome to Nix version 2.4pre20210326_dd77f71. Type :? for help.

nix-repl> builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
error: invalid regular expression '^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)'

       at «string»:1:1:

            1| builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
             | ^
            2|

x86_64-linux

[x86_64-linux] $ nix repl
Welcome to Nix version 2.3.10. Type :? for help.

nix-repl> builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
null

nix-repl> builtins.currentSystem
"x86_64-linux"


[x86_64-linux] $ nix repl
Welcome to Nix version 2.4pre20210326_dd77f71. Type :? for help.

nix-repl> builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
null

aarch64-linux

[aarch64-linux] $ nix repl
Welcome to Nix version 2.4pre20210326_dd77f71. Type :? for help.

nix-repl>  builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
null

x86_64-darwin

[x86_64-darwin]$ nix repl
Welcome to Nix version 2.3.10. Type :? for help.

nix-repl> builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
[ ]

Steps To Reproduce

1. On a aarch64-darwin system open the nix repl and try to eval builtins.match "^.*?(egg|tar|tar.bz2|tar.gz|tar.lz|tar.lzma|tar.xz|tbz|tgz|tlz|txz|whl|zip)" "asd"
2. You should see an invalid regex error now.

Expected behavior

The behavior should be the same across all the platforms.

nix-env --version output

nix-env (Nix) 2.4pre20210326_dd77f71

[vcunat]

I've already heard of some practical cases where x86_64-darwin was stricter than x86_64-linux. Perhaps this report belongs in here? #1537

I certainly agree that the differences are unfortunate. Maybe we could at least have ofborg check evaluation on multiple platforms or something.

[andir]

Good find! I haven't seen that issue when I searched. The interesting piece is that depending on architecture I am compiling for it works on darwin. Perhaps we are using different c++ library implementation.

[domenkozar]

Isn't this an upstream issue of the regex lib?

[stale]

I marked this as stale due to inactivity. → More info

[kamadorueda]

I also found this problem on x86_64-darwin, and the fix was removing the ? in the wildcard .*?: on-nix/python@4d3d14f

Seems that adding or not the ? yields the same results (at least in my case), with the difference that it does not break

[alex-ameen]

I got bitten by this same issue this week. It was driving me absolutely nuts.

I finally decided I wasn't crazy after running similar side by side comparisons.

At the very least a note should be added to the docs.

Related issue is x{0,} not working. Basically you can't do non-greedy matches on Darwin.

[domenkozar]

@toonn could you take a look at this one?