Warning about package lock violation during compile-time

[Stas Boukarev]

Doing COMPILE-FILE on something like "(defun car (x) x)" doesn't warn about package lock violation, the error will be signaled only when loading the produced file.

I got bitten by it when compiling (defun file-length (file) (file-length file)), because it signals its own warning regarding recursive definition of a known function. And the full warning prevented Slime from loading the fasl (failure-p was set to T on compile-file) and I didn't see the package lock violation error.

[Attila Lendvai (attila-lendvai)]

this sounds at least related to https://bugs.launchpad.net/sbcl/+bug/675633 (a wish to land compile-time errors in the debugger instead of turning them into compiler error messages).

[Alastair Bridgewater (alastair-bridgewater)]

It's not really related because the package lock isn't violated until LOAD-time, thus there /is/ no error at compile time.

[Nikodemus Siivola (nikodemus)]

A full warnings is probably too much, but a style-warning certainly sounds appropriate.

[Nikodemus Siivola (nikodemus)]

There are actually valid reasons to make it a full error:

1. Existing inline definitions can get frobbed.

2. The name becomes proclaimed as a function name, if it wasn't one before.

[Nikodemus Siivola (nikodemus)]

In 1.0.47.21.