Custom Query (1630 matches)

bloom-filter's test suite shows some errors on 2012-04-30: ​http://tests.call-cc.org/master/linux/x86/2012/04/30/salmonella-report/test/bloom-filter.html

The day before tests were working: ​http://tests.call-cc.org/master/linux/x86/2012/04/29/salmonella-report/test/bloom-filter.html

See the documentation. The "charplot" egg depends on this. We can change srfi-63 to export the uppercase identifiers but should then change charplot accordingly and update the documentation

We recently ran salmonella with the -specialize option (which implies the -scrutinize option), and it turns out that the tuples egg raises quite a few warnings when it's being compiled: ​http://parenteses.org/mario/misc/specialize-report/install/tuples.html The skiplists egg has the same problem: ​http://parenteses.org/mario/misc/specialize-report/install/skiplists.html

I attempted to track down the cause, but proved to be extremely difficult. It looks like it's not tuples itself that's in the wrong, but its use of the contracts egg is causing the trouble. When I compile the test suite of contracts with -specialize I also see a few warnings.

In the end I was able to reduce it to this program that fails:

(use contracts)

(define-with-contract (single xpr)
  (range (even? result))
  (lambda (sel) (print xpr)))

This is the warning I see:

$ csc -specialize contracts-test.scm
Warning: at toplevel:
  (contracts-test.scm:4) in procedure call to `g108109', expected 1 argument, but was given 0 arguments

When running with -debug 2, I see that single is a procedure which receives one "proc" argument, but it is called with zero arguments. Oddly enough, the code doesn't fail immediately. I really don't understand where it's going wrong, but the macros that apply here are quite complex! They're probably generating some dead code that is wrong even though it never gets called, and that's why the compiler still flags it.