diff options
author | sterni <sternenseemann@systemli.org> | 2023-06-13T13·15+0200 |
---|---|---|
committer | clbot <clbot@tvl.fyi> | 2023-06-15T11·01+0000 |
commit | 0005737f110aa7667aec3ca3bb5bc7d4907a8664 (patch) | |
tree | 9c7b46cfb23ea7ae3774ec69db942f63126eeb8e /nix/sparseTree/default.nix | |
parent | 9c7d1361c5a049634200c0153d8dcfec6931e89a (diff) |
fix(tvix/eval): make tvix display values like nix-instantiate(1) r/6308
In order for the test suite we have currently to be comparable to C++ Nix, we need to display values in the same way. This was largely the case except in some weird cases. * <CODE> for thunks and <CYCLE> for repeated thunks (?) are already in use. <CODE> formatting is tested by the oracle test suite already. * Instead of lambda, we need to use <LAMBDA> * <<primop>> and <<primop-app>> (a formatting C++ Nix uses nowhere) now are <PRIMOP> and <PRIMOP-APP>. We'll probably want to have a fancier display of values (in a separate trait) down the line. This could be used for interactive usage, e.g. the REPL or a potential debugger. There is a peculiarity with C++ Nix 2.3 formatting primops: import is considered a <<PRIMOP-APP>>, since it is internally implemented by means of scopedImport. This implementation detail no longer leaks in C++ Nix 2.13 nor in Tvix. <CYCLE> display is untested at the moment, since we exhibit a discrepancy to C++ Nix 2.3. Our current detection is more similar to C++ Nix 2.13—luckily it is also the more consistent of the two. See also b/245. Change-Id: I1d534434b02e470bf5475b3758920ea81e3420dc Reviewed-on: https://cl.tvl.fyi/c/depot/+/8760 Reviewed-by: tazjin <tazjin@tvl.su> Autosubmit: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
Diffstat (limited to 'nix/sparseTree/default.nix')
0 files changed, 0 insertions, 0 deletions