Remove file-level attributes disabling warning 4 in Flambda 2 #2314
Conversation
mshinwell
force-pushed
the
flambda2-warning-4
branch
from
df4c8b8 to
fc39deb
Compare
There was a problem hiding this comment.
I think the PR is correct, but I'm not very enthusiastic. In my opinion it makes the code harder to read because it's less obvious that the expanded case is a default case. Plus a few instances are really hard to read (a match in a when clause, for instance.)
| | Value_approximation.Value_symbol s -> Simple.symbol s | ||
| | _ -> simple_var | ||
| | Value_symbol s -> Simple.symbol s | ||
| | Value_int i -> Simple.const_int i |
There was a problem hiding this comment.
I'm not sure about this addition. We're creating a simple for the main module block, if it is an integer we have a serious problem.
There was a problem hiding this comment.
Hmm, maybe some of these should actually be errors indeed...
|
We will leave out the changes to |
This removes the file-level attributes in the main part of the Flambda 2 codebase that disable the fragile-match warning. In the majority of cases I have made patterns exhaustive rather than just pushing the attributes down onto particular matches. This has cleaned the provers up somewhat; even if they are all a bit verbose, they are at least now consistent.
I made a tiny semantic change in
Closure_conversionalong the way to propagate information from aValue_intapproximation.Based on #2312