Faré wrote:
The results are very encouraging: only 4 red things, that look fixable.
At the risk of being a party pooper, I don't think that these results, while encouraging, are measuring what needs to be measured.
What needs to be measured is the extent to which this change breaks systems that *use* these libraries, not whether the libraries in and of themselves fail to build.
While I am very grateful to cl-test-grid, we need to be conscious of its limitations. One of the things to which it is blind (in its current form) is the extent to which a change breaks a client system that is built out of the libraries in quicklisp.
So if I have an application that is built on a library that destructively modifies the default readtable, and I rely upon that modification, then my application could easily be broken, even if the cl-test-grid says that the library is fine.
The readtable is an (implicit) part of the library's API, and we are not measuring changes in that API. Because of the implicit nature of the API, measurement is even more difficult than it might otherwise be.
[Yes, I concede that this is not a *good* API, but ASDF is not "lint"; it's "make".]
This is one of the reasons that I feel cautious about introducing this change, and want 'strict mode' to be the optional case, not the default one.
Best,
r